--- title: "You Built a Brilliant AI Agent — Now Get 300 People to Use It" episode: 105 podcast: "The LeanScale Podcast" publisher: "LeanScale" guest: "Christopher O'Donnell" guest_title: "Founder & CEO" date_published: 2026-08-27 date_modified: 2026-09-03 duration: 01:11:04 word_count: 11375 topics: ["ai-in-gtm", "revenue-operations", "gtm-strategy", "sales-leadership"] canonical_url: https://www.leanscale.team/knowledge/podcast/christopher-odonnell-day-ai-single-player-multiplayer-agents/ source: "LeanScale Knowledge Hub — https://www.leanscale.team/knowledge" license: "Free to quote and cite with attribution to The LeanScale Podcast." --- # You Built a Brilliant AI Agent — Now Get 300 People to Use It _Day AI's Christopher O'Donnell on the folder every RevOps team is quietly building, and why multiplayer mode barely works_ **Episode 105 · The LeanScale Podcast** Christopher O'Donnell, Founder & CEO (Day AI) · Hosted by Anthony Enrico Published August 27, 2026 · Updated September 3, 2026 · 01:11:04 Canonical: https://www.leanscale.team/knowledge/podcast/christopher-odonnell-day-ai-single-player-multiplayer-agents/ **Topics:** AI in GTM · Revenue Operations · GTM Strategy · Sales Leadership ## Executive summary Building an AI agent that works is the easy part now. Christopher O'Donnell — a decade as HubSpot's Chief Product Officer, now founder and CEO of Day AI — opens with an observation about a folder. He has not met a RevOps person in sixty or ninety days without one: ICP and pipeline definitions at CRO-level detail, internal decisions rather than to-dos, Salesforce exports, Granola notes, Gong transcripts, enablement material. Mash it together in Claude Code and it works immediately. Then you try to get three hundred people to use it, and it hits a wall. Two things make that folder valuable. The first is internal meeting capture rather than customer calls — the decisions that evaporate, where companies spend enormous time in meetings and nobody knows what happened to what was decided. The second is documentation as a compounding asset. RevOps has always been the go-ask-Kevin role, because there are always caveats behind why a field or a Looker projection works a certain way. Christopher's test: he has never read the file, but if he were hit by a bus you could read it and have 100% of the context. The wall is multiplayer. Anthony's framing is that everybody's Claude account is the new Excel spreadsheet — individual formulas, individual data, people arriving at meetings with different numbers. SaaS solved that with a harness and governance; nothing equivalent exists for agents. Christopher's evidence is concrete: people abandon Salesforce MCPs, including, he says, Claude's own creators, who type deal notes into a browser tab. Founder god-mode agents have been spun down. And the central-graph approach was abandoned because it becomes a vendor-scale ingestion problem, some sources like Gmail you will not do internally, and the data needs permissioning. Anthony offers the schoolyard-fence study as the reframe: without a fence children stay close to the building, with one they range to the boundary. Boundaries expand what reps do. That matters because salespeople reply — they will tell the agent to send it Wednesday before the pipeline meeting, to triple-check nothing can be advanced, to give them something smart to say. And they are within their rights, because we pay salespeople to be themselves within a process. The answer Christopher describes is a deploy button: push central skills into personal harnesses and update them without destroying accumulated personalisation. The organising metaphor is the virtual employee — a name, a job description, skills that run on a schedule or trigger, and capability tiers, rolled out like a new hire class. It maps to how people already think about org design, and it makes a thousand-rep behaviour change one act of change management. Anthony forecasts an AI Operations team to manage which model each agent needs once the subsidy ends. The closing stretch covers OTEs doubling, managers going from seven reports to fourteen, sellers absorbing marketing — and then atrophy. Christopher's honest answer is that software engineering has had this magic for a while and nobody works less; the tendency is five times the work in 140% of the time. His North Star is twice the productivity in 80% of the hours, and his own habit is writing the twelfth draft himself on a blank page. ## Key takeaways 1. **Every RevOps person now has the folder** — Christopher has not met one without it in sixty to ninety days. Inside: ICP and pipeline definitions at CRO-level detail, documented decisions rather than to-dos, Salesforce exports, Granola notes, Gong transcripts, wiki pages, enablement material — mashed together in Claude Code. _Why it matters:_ The moment people start doing this they realise it works. RevOps curiosity then takes it to strategic altitude fast: what leverage can this give the company, not just what admin can it remove. _For:_ RevOps Leaders, Revenue Executives, Founders 2. **Internal meeting capture is the breakthrough, not customer calls** — Companies spend enormous time in meetings and genuinely do not know what happened to the decisions made in them. Christopher describes going into a well-known company's sales team and having exactly that surface as the real problem. Anthony's addition: COVID normalised recording and the expectation survived the return to the office, so the transcript archive predates the models that can now use it. _Why it matters:_ Customer-call capture is well served already. The decisions that evaporate between internal meetings are the unclaimed asset. He runs a services business by monitoring Slack channels and the transcript repo rather than asking people for updates. _For:_ RevOps Leaders, Revenue Executives 3. **Documentation in plain English becomes a compounding enterprise asset** — RevOps has always been the go-ask-Kevin role, because there are always caveats behind why a field or a Looker projection works a certain way. LLMs handle that nuance alongside quantitative data, so the asterisks move out of one person's head. _Why it matters:_ Christopher's test: he has never read the file, but if he were hit by a bus you could read it and have 100% of the context. In Looker, Tableau, Snowflake or Salesforce there was never a place or reason to write that down. _For:_ RevOps Leaders, Revenue Executives, Founders 4. **Your Claude account is the new Excel spreadsheet** — Anthony's framing: everyone built their own formulas and databases, then arrived at meetings with different numbers. SaaS fixed it with a harness, standardised tables and governance. Now everyone has their own repo, agents and skills again. _Why it matters:_ The problem appears the moment you want a hundred salespeople using the same thing the same way, with defined room to customise. That is a governance problem, not a capability one. _For:_ RevOps Leaders, Revenue Executives, Founders 5. **People abandon MCPs and central graphs — including at Claude's own maker** — Christopher says Claude's creators gave up on the Salesforce MCP internally and open a browser tab to type deal notes by hand. The central-graph approach people built earlier in the year has also been abandoned: it becomes a vendor-level ingestion problem, some sources like Gmail you will not do internally, and the data needs permissioning. The same fate met founder agent stacks: every founder Christopher spoke to through the spring was building one, often named Atlas or Sauron, one spending $300–500 a day on compute. Almost all were shut down or kept private. _Why it matters:_ Without the right binding into whatever harness you use, it simply does not work — however reasonable the request. Asking an agent to find every moment a gatekeeper became a champion falls over today. The moment you push it to other people it becomes a product — and these founders are engineers who forget reps are not technical. _For:_ RevOps Leaders, Revenue Executives 6. **Governance and permissioning are the unsexy things that let you scale** — Christopher compares the topic to discussing curfew with teenage kids. But permissioning natively — your agents see everything I do, I see your emails only when relevant to a deal, no Chrome extension required, with the AI managing it and humans as final judge — is table stakes. _Why it matters:_ Anthony's version: a founder can run agents loose with god-mode access, but agents built for an architect or engineer should not reach financial data or private messages. You cannot scale without it. _For:_ RevOps Leaders, Revenue Executives, Founders 7. **The schoolyard fence: boundaries expand what people do** — Anthony cites a study where children in an unfenced schoolyard stayed close to the building, while children with a fence ranged all the way to it. Boundaries produced more exploration, not less. _Why it matters:_ Guardrails give a sales team the confidence to use AI meaningfully rather than restricting them. Christopher had not heard the study and immediately believed it. _For:_ Revenue Executives, Sales Leaders, RevOps Leaders 8. **Salespeople reply to agents, and they are within their rights** — Send a rep a Slack message from an agent and they answer — what about this deal, that is not a deal, actually it is because someone is on vacation. They will also say send it Wednesday before my pipeline meeting instead of daily, triple-check nothing can be advanced, give me something smart to say. _Why it matters:_ These cannot be fire-and-forget webhooks; you need continuable conversation across platforms. And we pay salespeople to be themselves within a process, which is a paradox no agentic coding tool comes near solving. _For:_ Sales Leaders, Revenue Executives, RevOps Leaders 9. **The deploy button: central skills into personal harnesses** — Push what you want centrally controlled out like a Netlify or Vercel deploy, but have it live inside each person's own harness — then update those over time without destroying the personalisation and preferences everyone has accumulated. _Why it matters:_ Christopher expects this to be obvious within a year and says it is only just occurring to people now. It is the specific hard problem between a working prototype and team-wide rollout. _For:_ RevOps Leaders, Revenue Executives, Founders 10. **The virtual employee model is how agents get adopted** — Give an agent a name, a job description, skills that run on a schedule or a trigger, and capability tiers — then roll them out like a new hire class. Christopher has an agent called Chloe who owns podcast partnerships and a chief-of-staff agent that tells him which customer to write to. _Why it matters:_ It maps onto how humans already think about org design, so everyone understands it immediately. Unlike software-engineering agent definitions, these have to accept feedback and be coached rather than being universal. _For:_ Revenue Executives, RevOps Leaders, Founders 11. **Changing a thousand reps' behaviour as one act of change management** — The CEO wants everyone selling the new product. Review the last hundred meetings and email threads, identify where it should have come up, derive how to recognise that moment, review the automation at a granular level, deploy the edits — preserving each person's personalisation. _Why it matters:_ The number of people stops mattering. Christopher's framing: it does not matter whether it is a hundred or a thousand, it happens as one change under the covers. _For:_ Revenue Executives, RevOps Leaders, Sales Leaders 12. **An AI Operations team is coming, driven by cost** — Anthony's forecast: the subsidy ends, and someone has to decide which agent needs which model rather than running frontier models for everything. The costs are real and in some cases comparable to a full-time person. _Why it matters:_ Organising agents as roles gives you the framework to answer which model and which access level each one needs. Christopher's counterpart is tiering, so capability and cost scale together. _For:_ RevOps Leaders, Revenue Executives, Founders 13. **OTEs double, managers get twice the reports, sellers become marketers** — Christopher's finance plan has OTE doubling over 24 months — a rep on 200k making 400k at the same level, because an inside rep doing 2.5 quality calls a day will happily do four. Sales managers go from seven direct reports to fourteen. And senior sellers absorb marketing: a rep with a book of business can make the landing page, the ebook, the webinar. _Why it matters:_ The buyer experience becomes asking Claude or Perplexity a question, seeing a named author on the cited content, and working with that person throughout — what enterprise buyers of Coca-Cola have always had. _For:_ Revenue Executives, Sales Leaders, Marketing Leaders 14. **The atrophy question has no comfortable answer yet** — Anthony's framing: we are mathematicians who just got handed calculators — we did the work manually so we know how deep things should go. What about people who never did? Christopher calls it genuinely brilliant and outside a one-year horizon. _Why it matters:_ His evidence is that software engineering has had this magic for a while and nobody works less. The tendency is five times the work in 140% of the time, and he expects people who went deep early to burn out. _For:_ Founders, Revenue Executives, RevOps Leaders 15. **Twice as productive in 80% of the hours, and write the twelfth draft yourself** — Christopher's stated North Star is a salesperson working 80% of today's hours, with all of it value-add, making twice as much. His own practice: eleven versions of a memo shaped with Claude, then a blank Google Doc for the twelfth, written top to bottom with nothing open. _Why it matters:_ The first eleven versions turned out to be there to tease his imagination. Anthony's parallel is the gym — the industrial era did fitness through work, so the knowledge era may need deliberate cognitive exercise. _For:_ Founders, Revenue Executives ## Frameworks ### Single-Player vs Multiplayer AI (12:11) **Definition:** The single-player experience — one person with a folder, a harness and full access — is already excellent. The multiplayer experience, requiring governance, permissioning and personalisation at scale, barely works. Christopher calls this the single biggest topic in go-to-market right now. The gap is why founder god-mode agents get spun down and why central-graph projects were abandoned mid-year. ### Your Claude Account Is the New Excel Spreadsheet (12:11) **Definition:** Individual agents and skills recreate the pre-SaaS spreadsheet problem: everyone with their own formulas, their own data, arriving at meetings with different numbers. Anthony's framing. SaaS solved it with a harness, standardised data tables and governance that let a large organisation use spreadsheet-grade power collectively. No equivalent exists yet for agents. ### The Schoolyard Fence (23:06) **Definition:** Children in an unfenced schoolyard stay close to the building; children with a fence range all the way to the boundary. Constraints expand exploration rather than limiting it. Anthony's reframe of why governance is not the enemy of adoption. Give a sales team AI with the right guardrails and they use it further and more confidently than without. ### The Deploy Button Into Personal Harnesses (27:39) **Definition:** Push centrally-controlled skills out like a web deployment, but have them live inside each person's own harness, so central updates land without overwriting accumulated personalisation. The specific hard problem between a working prototype and a team-wide rollout. Christopher expects it to be obvious within a year and says people are only starting to see it now. ### The Virtual Employee (30:53) **Definition:** Agents with a name, a job description, a defined skill set that can run on a schedule or a trigger, and capability tiers — rolled out to managers like a new hire class. Unlike software-engineering agent definitions, which are universal, these must accept feedback and be coached. Its advantage is that it maps to how people already think about org design, so adoption is immediate. ### Behaviour Change as One Act (35:32) **Definition:** Review the actual day in the life of sellers across the last hundred meetings and threads, identify the missed moments, derive the recognition insight, review the automation granularly, then deploy edits across every personal harness at once. Christopher's worked example is a CEO wanting the new product sold. The count of people stops being a variable — a hundred or a thousand is the same operation. ## Quotes _Speakers inferred from an undiarized transcript — verify before attributing._ > "Every single person I talk to that works in this field, or anywhere close to it, go-to-market engineering, or CRM admin, anything. People have started collating documents around what their company is trying to do." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (00:20) > "There may be RevOps people out there that don't have one of these folders on their laptops. I have not talked to one in the last 60, 90 days." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (02:10) > "The minute you start doing that, you realize, "Oh, this works. There is something here."" > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (04:01) > "They've been unlocked in a way where they've been held back for so long." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (05:22) > "A big breakthrough for us has been understanding the value of internal meeting capture vis-a-vis go-to-market. It's not just about the customer calls." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (06:11) > "I've never read the file that that's in, but if I get hit by a bus, you could read that file and you would know exactly 100% of the context that's behind it." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (08:26) > "We're just sitting on absolute gold in our transcripts for meetings." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (09:29) > "Everybody's cloud account is basically their Excel spreadsheets. This was the old problem." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (12:11) > "The creators of Clod internally have given up using Salesforce MCP, and they open up a tab in their browser to just enter the deal notes." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (14:09) > "If you don't have the right kind of binding into your harness, whatever it is, it does not actually work." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (14:45) > "I talked to those same people, and they have given up on that." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (16:28) > "You need to have permissioning on the data. I mean, it's not a fun thing to talk about. It's a little bit like discussing curfew with your teenage kids or something." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (16:55) > "You cannot scale unless you have the governance and permissions." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (18:13) > "Every single one of them that I've talked to has given up on this and spun them down or kept them just for themselves." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (19:45) > "These are salespeople. These are not technical folks. They're going to interact with it the way they interact with it. And you're not going to change that." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (20:21) > "They're within their rights to tell the agent what they want, especially salespeople, where we give salespeople a lot of room to be themselves." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (21:54) > "Having boundaries actually expands how far people will go." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (22:59) > "I just don't have these CRM data entry blocks on my calendar anymore. And I'm never going to in my career again." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (24:29) > "How do you from your folder, push, almost like a Vercel or Netlify, like deploy button, push skills out, but then they live within a personal harness." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (27:39) > "The virtual employee kind of concept is correct." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (30:53) > "You roll these agents out as though you were introducing a new hire class." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (33:24) > "Organizing them into roles and jobs, that is a container for these things." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (37:58) > "Actually like change the behavior of a group of a hundred people, a thousand people, like it literally doesn't matter, Anthony, how many people there are." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (37:27) > "I'm forecasting there's going to be a team that's called AI operations within every company." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (41:48) > "There are real hard costs to leveraging AI. In some cases, it's comparable to bringing on a full-time person." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (42:29) > "I think OTEs for sales reps are going to double. My finance plan has OTE doubling over 24 months here." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (48:17) > "Sales managers will go from seven direct reports to twice that." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (49:00) > "I see a senior seller, for example, starting to do a lot more marketing as we specialize." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (49:48) > "You're a rep, you have a book of business. Now maybe you know this type of company within that book of business. That's your whole job. You can make a landing page, dude." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (50:50) > "Work is getting really fun, I will say. It's getting really fun that we're focused on the creative side 100% of the time." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (53:53) > "We are mathematicians that just got handed calculators." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (59:57) > "It has not caused people to work less. It has not caused people to be less stressed." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (1:00:47) > "The tendency is to do five times as much work in 140% of the time." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (1:01:32) > "You're leaving Thanksgiving dinner halfway through to check on your quad code processes and keep them going and unblock them." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (1:02:27) > "I want to see a salesperson work 80% of the hours they work today have all of that time be what they consider their value add and make twice as much." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (1:03:02) > "If you don't actually know the subject, then you don't have taste for what good looks like in this area." > > — Anthony Enrico, The LeanScale Podcast Ep. 105 (1:08:01) > "And then the 12th version, I opened up Google docs, a blank document, and I looked at nothing. And I just sat and I just wrote." > > — Christopher O'Donnell, The LeanScale Podcast Ep. 105 (1:08:51) ## Practical advice by role ### RevOps Leaders - Build the folder deliberately: ICP and pipeline definitions at CRO-level detail, documented decisions rather than to-dos, and the caveats that only live in your head. - Capture internal meetings, not just customer calls — that is where decisions currently evaporate. - Write the asterisks in plain English so someone else has 100% of the context without you. - Do not attempt a central graph you ingest yourself; it becomes a vendor problem, some sources are off limits, and it still needs permissioning. - Design for continuable conversation, not fire-and-forget notifications — reps will reply, and they should. ### Revenue Executives - Treat governance and permissioning as the enabler of scale rather than a compliance chore. - Roll agents out as a new hire class with names, job descriptions, skills and tiers — it maps to how people already think. - Plan for an AI Operations function to decide which model and access level each agent needs once subsidised pricing ends. - Prepare for the shape change: OTEs roughly doubling, managers carrying twice the reports, senior sellers absorbing marketing work. - Set the productivity target as more value-add in fewer hours rather than more output in more hours. ### Sales Leaders - Expect reps to negotiate with the agent — cadence, tone, what to triple-check — and treat that as legitimate rather than as non-compliance. - Give guardrails so reps range further, not less; the fence is what produces exploration. - Eliminate CRM data entry first — it is the change reps notice within days. - Support managers explicitly through the doubling of span; many are new to management and this is where the strain lands. ### Founders - Assume your god-mode agent will not survive contact with non-technical users; the moment you push it out it becomes a product. - Solve the deploy-to-personal-harness problem before trying to standardise anything across a team. - Watch for the five-times-work-in-140%-of-the-time failure mode in yourself and the people who went deep earliest. - Keep one thing you do unaided — Christopher writes the twelfth draft on a blank page after eleven AI-assisted versions. ## AI takeaways **Thesis:** The whole conversation is a single argument: capability is solved and distribution is not. Any RevOps person can build something genuinely useful in a day, and almost nobody can get three hundred people using it. What stands between them is governance, permissioning, and the ability to push central updates into personalised harnesses — which is why founder god-mode agents get spun down and central-graph projects were abandoned mid-year. - **** — - **** — - **** — - **** — - **** — - **** — **Agent & automation ideas** - An agent updating CRM records on reps' behalf and showing its reasoning for each opportunity or field change. - Meeting prep that researches similar stuck deals and how they were unlocked, with the rep able to tune tone and directness. - A named partnerships agent that finds opportunities, explains why they fit, and drafts the outreach — Christopher's Chloe. - A chief-of-staff agent surfacing which customer or colleague to contact and why. - An internal-decision capture agent recording what was decided in meetings rather than what the action items were. - A behaviour-change agent reviewing the last hundred meetings to find where a new product should have been mentioned, then deploying the edits. ## Operations takeaways ### Revenue operations - **.** - **.** - **.** - **.** - **.** ### Pipeline & marketing ops - **.** - **.** - **.** - **.** ### Customer operations - **.** - **.** - **.** ## Metrics mentioned | Value | Metric | Context | | --- | --- | --- | | None in 60–90 days | RevOps people without the folder | Christopher's count of how universal the pattern has become among people he speaks to. | | $300–500 per day | Founder agent compute spend | One founder's Google Cloud bill running a personal god-mode agent stack, which was subsequently spun down. | | Doubling over 24 months | Forecast OTE change | Christopher's own finance plan — a rep on $200k reaching $400k at the same level through efficiency rather than promotion. | | 7 → 14 direct reports | Sales manager span of control | What Christopher describes as approaching consensus forecast, and the group he thinks will feel the strain most. | | 2–2.5 quality calls a day → 4 | Rep call capacity | The efficiency change underlying the OTE forecast, once reps stop doing work they should not be doing. | | +30% to +60% | Active opportunities per rep | Attributed to Mark Roberge's analysis, varying by company, as the input to higher OTEs. | | 2x output in 80% of hours | Productivity North Star | Christopher's stated internal goal, against the observed tendency of five times the work in 140% of the time. | | 11 AI-assisted, 12th by hand | Drafts before writing unaided | His own practice for a memo — the assisted versions served to tease imagination rather than produce the artefact. | ## Entities mentioned - **Day AI** (company) — Christopher's company, building an AI-native agentic GTM stack with customer memory, governance and permissioning, backed by Sequoia. Ships a Claude Code harness and a GTM brain template repo that Claude can read directly from a public GitHub URL, which he describes as tens of thousands of dollars of onboarding sitting in a file. · https://www.leanscale.team/knowledge/company/day-ai/ - **HubSpot** (company) — Where Christopher spent a decade as Chief Product Officer, building the Marketing, Sales, Service and Data Hubs from scratch. Anthony references it as his alma mater when discussing content strategy. · https://www.leanscale.team/knowledge/company/hubspot/ - **Sequoia Capital** (company) — Backer of Day AI, named in the introduction to the agentic GTM stack Christopher is building. · https://www.leanscale.team/knowledge/company/sequoia/ - **Netlify** (company) — Used as the deploy-button analogy for pushing agent skills into personal harnesses. Christopher notes Day AI invested at Series C and names people there; Anthony confirms LeanScale deploys to it, and jokes that Claude keeps suggesting Vercel instead. · https://www.leanscale.team/knowledge/company/netlify/ - **LeanScale** (company) — Anthony's firm. He describes running a people-heavy GTM ops agency by monitoring Slack channels and the transcript repo rather than asking for updates, and estimates the team is at least twice as productive as before. · https://www.leanscale.team/knowledge/company/leanscale/ - **Christopher O'Donnell** (person, guest) — · https://www.leanscale.team/knowledge/guest/christopher-odonnell/ - **Anthony Enrico** (person, host) — Co-founder of LeanScale and host of The LeanScale Podcast. · https://www.leanscale.team/knowledge/guest/anthony-enrico/ - **Claude Code** (tool, AI Dev Tool) — The harness Christopher says RevOps people are overwhelmingly using for the folder, because it is so good for mashing raw material together. Day AI ships a harness for it, and the worked thousand-rep change begins there. - **Salesforce** (tool, CRM) — The MCP example of single-player AI falling over — Christopher says Claude's own creators abandoned it internally and type deal notes into a browser tab by hand. - **Gong** (tool, Revenue Intelligence) — Named repeatedly as raw material for the folder and as a connectable data source, including in the example of finding every moment a gatekeeper became a champion. - **Granola** (tool, Meeting Notes) — Cited alongside Gong and Salesforce exports as the meeting-notes material RevOps people are collating into the folder. - **Looker** (tool, BI / Data Modeling) — Used as the example of a system where a projection carries context nobody wrote down — the caveats that historically lived only in a RevOps person's head. - **Slack** (tool, Team Messaging) — Both a data source needing permissioning — private conversations, changed channel permissions — and the delivery surface where reps reply to agents rather than accepting fire-and-forget notifications. - **Cursor** (tool, AI Dev Tool) — Named with other early AI apps that stubbed their toes on pricing and usage limits, because there was no intuitive concept for running out — a problem Christopher argues the virtual-employee tiering model solves. - **Perplexity** (tool, AI Search / Research) — Named with Claude and ChatGPT in the buyer-experience forecast: ask a question, get an answer, follow a citation to content with a named author, then work with that person throughout. ## FAQ **Q: What is the folder every RevOps team is building?** A: A collection of documents about what the company is actually trying to do: ICP definitions, pipeline definitions at CRO-level detail, internal decisions rather than to-dos, plus raw material like Salesforce exports, Granola notes, Gong transcripts, wiki and Notion pages and enablement content. Christopher O'Donnell says he has not spoken to a RevOps person without one in the last sixty to ninety days. Mashed together in Claude Code it works immediately, and what makes it powerful is that RevOps curiosity quickly takes it to strategic altitude — what leverage can this give the company, rather than what admin can it remove. **Q: Why is internal meeting capture more valuable than customer call capture?** A: Because customer calls are already well served, while internal decisions evaporate. Companies spend enormous amounts of time in meetings and frequently do not know what happened to the decisions made in them — the "didn't we say we were going to…" problem. Christopher describes working with a well-known company's sales team where exactly this surfaced as the real gap. Plenty of tools extract action items from a meeting; far fewer capture the decisions and the reasoning behind them. **Q: Why do people abandon MCPs and self-built central data graphs?** A: MCPs for standard go-to-market apps often do not perform well enough in practice — Christopher says Claude's own creators gave up on the Salesforce MCP internally and type deal notes into a browser tab by hand. The central-graph approach, where someone ingests everything into Elasticsearch, TurboPuffer or Pinecone and puts an agent in front of it, has largely been abandoned for three reasons: it becomes a vendor-scale ingestion problem, some necessary sources such as Gmail are not realistically done in-house, and above all the data needs proper permissioning. **Q: Why is governance the thing that lets AI scale?** A: Because without it you cannot give anyone but yourself access. A founder with god-mode access is never blocked, but agents built for an architect or an engineer should not reach financial data or private messages. The model Christopher describes is native permissioning — your agents see everything relevant, I see your emails only when they relate to a deal, no browser extension required, with the AI managing the permissioning and humans as final judge. He compares the topic to discussing curfew with teenagers: not enjoyable, and unavoidable. **Q: What is the virtual employee model for rolling out agents?** A: Give each agent a name, a job description, a defined set of skills it is trained to perform, and the ability to run those on a schedule or in response to a trigger — then tier them so higher tiers do more work and have access to more powerful capabilities. Roll them out to managers as though introducing a new hire class. Christopher has an agent called Chloe who owns podcast partnerships and a chief-of-staff agent that tells him which customer to write to. The advantage is that it maps onto how people already think about organisational design, so essentially everyone understands it immediately. **Q: How do you change the behaviour of a thousand reps at once?** A: As one act of change management. Start in Claude Code by reviewing the actual day in the life of sellers — the last hundred meetings, email threads and opportunities — to identify where something should have happened and did not, such as a new product never being mentioned. Derive the insight for recognising that moment, review the existing automation at a granular level, then deploy the edits into every personal harness while preserving each person's accumulated personalisation and preferences. Christopher's point is that the number of people stops mattering: a hundred or a thousand is the same operation. **Q: Why do AI notifications to salespeople need to be conversations?** A: Because reps reply. Send a Slack message listing the deals that matter and a rep will answer — what about this one, that is not a deal, actually it is because someone is on vacation. They will also tell the agent to send it Wednesday before the pipeline meeting instead of every morning, to triple-check that nothing can be advanced further, and to give them something smart to say. Christopher's view is that they are entirely within their rights, because salespeople are paid to be themselves within a process — a paradox no agentic coding tool comes close to solving. **Q: How are go-to-market roles and pay expected to change?** A: Christopher's finance plan has on-target earnings doubling over 24 months — a rep earning $200k reaching $400k at the same level, because an inside rep managing two or two and a half quality calls a day will happily do four once freed of work they should not be doing. Sales managers are expected to go from around seven direct reports to fourteen, which he thinks is where the strain will land hardest. And senior sellers are likely to absorb marketing: a rep with a book of business can now produce the landing page, the ebook and the webinar themselves. **Q: Is there an atrophy risk from relying on AI?** A: Christopher calls the question genuinely brilliant and puts it beyond a one-year horizon. His evidence is that software engineering has had this level of assistance for a while and it has not made people work less or feel less stressed — the tendency is to do five times as much work in 140 per cent of the time, and he expects some people who went deep early to burn out. His stated goal instead is twice the productivity in eighty per cent of the hours, with all of that time being value-add. His own counter-practice is writing the twelfth draft of a document unaided on a blank page after eleven AI-assisted versions. ## Timeline - **00:00** — Cold open + intro - **01:47** — The one folder every RevOps team is quietly building - **05:45** — Internal meeting capture: the breakthrough hiding in your Zooms - **06:51** — "If I get hit by a bus": documentation as a compounding asset - **12:11** — Your Claude account is the new Excel spreadsheet - **13:28** — Where single-player AI falls over (and why MCPs get abandoned) - **16:55** — Governance and permissioning: the unsexy thing that lets you scale - **22:59** — The schoolyard fence: how boundaries expand what reps do - **27:39** — The deploy button: pushing skills into personal harnesses - **30:53** — The virtual employee: rolling out agents like a new hire class - **35:32** — Changing the behavior of 1,000 reps as one change - **41:48** — The coming "AI Operations" team - **46:37** — Where GTM is going: OTEs double, sellers become marketers - **59:37** — Mathematicians with calculators: the atrophy question - **1:08:31** — The 12th draft: writing the blank page yourself ## Related episodes - **Ep. 103: Why Your Reps Should Never Open the CRM Again** (Justin Lee (Superblocks)) — The same last-mile problem from the operator's side — god-mode prototypes that cannot be handed to reps until authorisation is scoped. - **Ep. 100: AI Ops: How We Run RevOps for 30 SaaS Companies at Once** (Jake Toepel (LeanScale)) — The AI Operations function Anthony forecasts here, already running across a portfolio. - **Ep. 98: AI-Native GTM: 3 Agent Plays and the Layer That Makes Them True** (Jake Toepel (LeanScale)) — The context-graph argument for why agent answers need an agreed model of the business beneath them. - **Ep. 102: I Don't Want Your Product. I Want Your Expertise.** (Noah Marks (GTM Council)) — The buyer-side view of the same moment — anyone can build, so what survives is expertise and governance. - **Ep. 93: How HubSpot Built a Media Empire** (Jonathan Hunt) — The content strategy behind Christopher's forecast that senior sellers absorb marketing work. · https://www.leanscale.team/knowledge/podcast/jonathan-hunt-hubspot-media-empire/ - **Ep. 99: Why Deflection Is the Wrong Way to Measure AI** (Dvir Ginzburg (Encore AI)) — The other integration-is-the-ceiling argument, from customer-facing rather than internal agents. ## Full transcript _Machine-transcribed and not diarized; speaker attribution is inferred._ _Transcript only, as a separate file: https://www.leanscale.team/knowledge/podcast/christopher-odonnell-day-ai-single-player-multiplayer-agents/transcript.md_ ### 00:00 — Cold open + intro **[0:00]** I do think we will see a lot of people from this era who went really deep early just run. **[0:08]** My guest today is Christopher O'Donnell, founder and CEO of Day AI, **[0:14]** where he's building sales agents with customer memory, an agentic GTM stack backed by Sequoia. **[0:20]** Every single person I talk to that works in this field, or anywhere close to it, go-to-market engineering, **[0:26]** or CRM admin, anything. People have started collating documents around what their company is trying to do. **[0:37]** It comes back, it's just like, this is what you need to do as a company. Like, great. **[0:43]** It's not to end, just a quick sidebar. Also, you were talking about speed to table stakes, **[0:49]** then Kimi coming out with a relevant model at a fraction of the cost. A couple months from now, **[0:55]** the capabilities that you'll be able to have access to at really, really affordable levels, **[1:02]** I don't know. And then whatever the next frontier model is, I can't even really imagine what that's **[1:06]** going to unlock. We are at a unique time where we've done all of the work in manual mode. **[1:12]** Are there any concerns around atrophy and the ability to leverage these models if you don't have **[1:21]** some baseline of skill as the foundation in the first place? I think it is a genuinely **[1:27]** brilliant question, and extremely relevant. This is outside of the one-year time frame. **[1:35]** I would say really the question. ### 01:47 — The one folder every RevOps team is quietly building **[1:47]** Christopher, you've said almost every RevOps person you talk to now has this one folder **[1:52]** that's become the center of their universe. Walk me through what's actually in it and when you **[1:57]** started noticing the pattern and problem. Yeah, I mean, first of all, thanks for having me. I've **[2:04]** really been looking forward to this, and it's great to see you again. There may be RevOps **[2:10]** people out there that don't have one of these folders on their laptops. I have not talked to **[2:17]** one in the last 60, 90 days. Every single person I talk to that works in this field, **[2:25]** or anywhere close to it, go to market engineering or CRM admin, anything, people have started **[2:33]** collating documents around what their company is trying to do. ICP definitions, pipeline definitions **[2:42]** at a really, really detailed level, like a CRO, sales ops-y kind of level, a CEO kind of level. **[2:51]** Just documenting everything, decisions, internal meetings, not just like to-dos. I mean, there are **[2:58]** a million apps that can tell you after a meeting what your to-dos are from the meeting, but real **[3:02]** decisions, and then starting to expand from there. A lot of RevOps people are very technical, **[3:09]** obviously. Maybe they're not software engineers. More and more of them that I run into have some **[3:15]** software engineering background, which I think is really cool for career paths and sales **[3:21]** productivity, certainly. But it doesn't matter anymore. You know what I mean? You can become **[3:26]** familiar with cloud code. I think particularly that seems to be really what people are using, **[3:34]** because it's so good for this. That basic recipe of, "Well, wait a second. I can take all of this **[3:42]** raw material that is sitting around a Salesforce export, and some granola notes, **[3:48]** and a Gong transcript, and a Wiki page, a Notion doc, and sales training stuff, **[3:55]** enablement stuff. I can just start mashing it up in cloud code. The minute you start doing that, **[4:01]** you realize, "Oh, this works. There is something here." Because these are RevOps people doing it, **[4:09]** I feel like their curiosity and creativity very quickly brings them to the strategic altitude. **[4:16]** What leverage can I offer the company here in terms of implementing big changes, big decisions, **[4:25]** in terms of automation, in terms of driving behavior, starting to really change the life **[4:33]** of a frontline seller out on the sales floor, incorporating marketing, giving feedback to **[4:38]** marketing, collaborating with product. It goes a bunch of different directions. Everybody seems **[4:43]** to have one with a similar kind of core. Yeah, and I think you're also dealing with **[4:49]** a group of people that are really tinkerers. They're used to working with a multitude of tools **[4:54]** and being flexible in the exact discipline, cross marketing, sales, CS. I know for sure, **[5:01]** I have turned my computer into the ultimate repo of all the things that I need, and almost becoming **[5:09]** a hoarder of knowledge and skills and things that I use to operate every day. It has opened up this **[5:17]** whole new world. I think, especially in RevOps, these people are fascinated by it, and they've **[5:22]** been unlocked in a way where they've been held back for so long. There are so many analyses **[5:30]** I've wanted to run. There are so many permutations of a plan I've wanted to model out, but it just **[5:36]** wasn't really practical before, and now the work almost seems unlimited. I totally relate to what ### 05:45 — Internal meeting capture: the breakthrough hiding in your Zooms **[5:45]** you're saying. These are folks who are patient, technical, in the room for lots of meaningful **[5:53]** conversations. There's a lot of content around them. If they're running granola notes or whatever it is, **[6:00]** they're exposed to a lot of kind of high stakes information and decisions, a lot of which **[6:05]** traditionally just totally gets lost. It's like, "Oh, didn't we say that we were going to ... " Well, **[6:11]** I guess I can't really remember, blah, blah, blah. A big breakthrough for us has been understanding **[6:17]** the value of internal meeting capture vis-a-vis go-to-market. It's not just about the customer **[6:23]** calls. It's about these internal meetings. We had one very well-known company in particular where we **[6:28]** went in to work with their sales team. The more we talk to them, they're like, "Oh, yeah. We spend **[6:37]** so much time in meetings, and we don't really know what happens with these decisions that we make." **[6:42]** It's like, "Oh, yeah. Right. Let's do something about that. Okay, cool." ### 06:51 — "If I get hit by a bus": documentation as a compounding asset **[6:51]** I think the other thing about RevOps people is this is one of those classic roles, like the **[6:57]** classic role of you would have to go ask Kevin. You would have to go ask Ariana, whoever, **[7:07]** if you want a certain report or if you want to drill into something or understand why something **[7:11]** is this way and not that way. Because as clean as the system gets, there are always caveats. **[7:18]** There's always context behind a field that gets added in a snowflake view or context behind why **[7:27]** a looker projection of the data is a particular way. The LLMs working so well with natural language **[7:37]** being able to flip back and forth between processing quantitative data and following **[7:45]** accounting for all of this asterisk, caveat, consideration, nuance type of stuff that only **[7:53]** lived in a RevOps person's brain. It's really nice for the RevOps person. I think it's also **[7:58]** really good for just enterprise value to have all of this stuff documented in plain English. **[8:05]** Even if you're doing this stuff, I do a lot of this where it's like, "Oh, that actually should **[8:13]** count as a pilot in that time period, even though there was a card in Stripe or whatever." **[8:22]** Document that somewhere, update your agents and skills, and make sure that that's always **[8:26]** explained and accounted for. I've never read the file that that's in, but if I get hit by a bus, **[8:35]** you could read that file and you would know exactly 100% of the context that's behind it. **[8:42]** In a system like Looker, Tableau, or Databricks, or Snowflake, whatever, Salesforce, **[8:49]** we're doing all of this data work and there's no place or reason really to document what it all **[8:57]** means. Now that we're in this world of like, "Oh no, we're just talking about it as we're working **[9:03]** with the hard quantitative data," you're also starting to get this compounding asset of what **[9:09]** everything is and why. Something too, I think I've talked about this before. I wonder if the moment **[9:19]** would be as big if the push to work remote from COVID didn't happen, because I've also realized **[9:29]** that we're just sitting on absolute gold in our transcripts for meetings. Now, at least if you're **[9:35]** in the B2B tech and SaaS space, most of your meetings are going to be recorded. You're going **[9:42]** to have a transcript for everything. Realizing how much value is there and then being able to **[9:49]** pipe that into the agents and skills that you've built has been an absolute game changer for me. **[9:55]** I'll give you an example. I run a GTM ops agency. We do GTM ops for companies. It's a very heavy **[10:03]** people process type of company, but I can keep track of everything going on because I can monitor **[10:10]** the Slack channels. I can monitor the transcript repo and I can see everything that's happening **[10:15]** with our customers and flag anything that's going on. All of that would have been, I don't even **[10:21]** think it would have been possible to do before AI. I think you make a great point. We had COVID, **[10:29]** we got very used to Zoom, and then we got very used to having transcripts and notes. **[10:37]** This was obviously well before these language models. As we returned to the office or partially **[10:43]** returned to the office, we kept that expectation and behavior to where the average team, if it's **[10:50]** just four people in a room ad hoc talking about something, somebody will capture it. You know what **[10:56]** I mean? It's like everything is recorded. What do you do with it? It's like you have this raw **[11:05]** material. It's a Slack message from a customer who reports a bug. It's a feature request from **[11:12]** a key account on a call that you record with Gong. It's the same feature request that another key **[11:19]** account said was make or break for them the week before. You have all of that. You have RevOps **[11:29]** people who are starting to go, "Well, wait a second. I can cook that together into these **[11:36]** incredible assets," but also it should be not just a brain. It should be like a whole organism that **[11:42]** is just moving and acting under my direction with the correct permissioning, the correct feedback **[11:51]** and eval, the correct ownership of that automation and who can personalize it and who can't. **[12:00]** That intuition is developing, I feel like right now. When you and I were talking about this, **[12:05]** I threw some of this by you and you were saying you were seeing a ton of it out there, right? ### 12:11 — Your Claude account is the new Excel spreadsheet **[12:11]** Oh, I think what's happening right now, so everybody's cloud account is basically **[12:17]** their Excel spreadsheets. This was the old problem. Everybody ran around with their Excel spreadsheets. **[12:22]** They built their own formulas. They had their own database of things that they thought were **[12:26]** important to them. Then people show up to a meeting with different spreadsheets and different **[12:31]** numbers and different things going on. What SAS did is it created that harness. It created that **[12:37]** governance. It created that standardization of those data tables that enabled a large organization **[12:46]** to take advantage of the power of something like a spreadsheet, but across the entire work. **[12:52]** Now it's happening. Everybody has their own repo, their own agents, their own skills. They're **[12:56]** running in their own unique way. They have their own nuances, but when you want to start using it **[13:02]** at scale, like, okay, I built this thing. That's really interesting and really helpful. Now I want **[13:07]** every single salesperson to use this thing in the exact same way or carve out what nuances **[13:13]** and regulations they're allowed to change and customize for them. That becomes a massive **[13:21]** problem. You're exactly right. I mean, for every Twitter article that you read where people are ### 13:28 — Where single-player AI falls over (and why MCPs get abandoned) **[13:28]** talking about the agents they've built for their company. You say to yourself, okay, well, what **[13:35]** happens when you need email data? What happens when you need data from private Slack conversations, **[13:45]** or if somebody changes the permission on a Slack channel? How do you get all of this stuff in a **[13:52]** format, first of all, that actually performs? Even just using single player, literally, **[14:01]** using Clod with MCPs for these kinds of standard go-to-market apps that we're talking about that **[14:09]** are well-built products that have been around for a long time. People give up. The creators of Clod **[14:20]** internally have given up using Salesforce MCP, and they open up a tab in their browser to just **[14:27]** enter the deal notes, which I find to be pretty hilarious that they're entering deal notes about **[14:32]** the god-tier AI demo that they just did for some prospect or something. It's like, okay, **[14:39]** better open up a new browser tab to salesforce.com and punch my notes in by hand. What? It's **[14:45]** ridiculous. But if you don't have the right kind of binding into your harness, whatever it is, **[14:54]** Clod code or a first-party web app like ours or Clod web, whatever, if it's not set up to actually **[15:00]** work, it does not actually work. If you say, go run across every gong call and all the Slack **[15:09]** conversations and find moments where the prospect really gave up on the deal or the gatekeeper **[15:18]** became a champion. I want to see all those moments and drill in and explore that. These **[15:23]** are reasonable things. Things that, by the way, anecdotally, I think six weeks ago, **[15:28]** if I got on a Zoom and somebody was trying to do that, I'd be like, oh, okay, you're a super early **[15:33]** adopter. And now I'm seeing people like, yeah, I am trying to do that. I have tried to see how the **[15:41]** new sales reps are doing compared to the tenured reps. That stuff just falls over. You can't **[15:48]** actually get it to work. Then you get into, okay, let's put all of our data into one place. **[15:57]** And folks, I think earlier this year were getting Elasticsearch or TurboPuffer or Pinecone, **[16:02]** whatever. And some really clueful person was maybe borrowed from engineering or something, **[16:09]** was ingesting all of that into some central graph or database or whatever and putting that **[16:15]** in front of the agent. And the problem with that is, sure, it's a lot of work. Some of it's not that **[16:20]** hard to do, actually. And people got a fair bit of the way. We're talking in mid-July, late July. **[16:28]** I talked to those same people, and they have given up on that. Partially because it turns **[16:35]** into a vendor-level problem to just ingest this data. Partially because they have realized they **[16:42]** need data sources, like, for example, Gmail, that you're just not going to do internally. You're just **[16:47]** not doing that, trust me. And then largely because you need to have permissioning on the data. I mean, ### 16:55 — Governance and permissioning: the unsexy thing that lets you scale **[16:55]** it's not a fun thing to talk about. You know what I mean? It's a little bit like discussing curfew **[17:00]** with your teenage kids or something. It's not the most fun thing to come into work and be excited **[17:06]** about. But we get pretty excited about it, just having worked in this stuff for so long. Doing **[17:11]** that natively so that if I come work for you, Anthony, maybe you can see all of my emails in **[17:18]** that graph. Your agents can see everything I'm doing. I can see emails from you, but only when **[17:26]** they're relevant to a deal. And you don't have to install a Chrome extension or do anything like **[17:32]** that. It's like the AI is managing all of that permissioning with us as the final judge. That **[17:38]** type of modern system, I think, it's just table stakes. Absolutely. Yeah, it won't be able to **[17:45]** scale without it. And it's easy for the people who are posting on X and whatever, or even people **[17:51]** like me, like, yeah, founder, CEO, I have God mode access to all of our data. So I can make any **[17:57]** agents I want and just have it run loose. But when I want to build agents that are specific to an **[18:03]** architect or engineer on the team, I don't need them to have access to our financial information. **[18:08]** I don't need them to have access to my emails or my private slacks. Those are things that **[18:13]** there's obviously lots of sensitive data. So we do, you cannot scale unless you have **[18:20]** the governance and permissions. And I really do think there's a lot of different aspects of this **[18:26]** wave compared to the SAS wave. But I do think SAS did bring in that standardization compliance and **[18:33]** governance that people could rely on and count on. And that is what enabled them to scale. So yeah, **[18:38]** it's not the sexiest topic, but it's necessary if you want to get the actual enterprise value **[18:44]** out of what people are building. A hundred percent, you know, and it's a lot of founders **[18:50]** that I talked to that, I mean, basically do every founder I talked to through the spring **[18:57]** was building and they always had a name for it, you know, Atlas or, you know, you know, **[19:04]** Sauron or whatever. I'm like, I don't know if you want to call it Sauron, you know, but, you know, **[19:09]** coming up with a personal agent and in the back of their heads, they're thinking, boy, **[19:15]** this is going to get so good that I can generalize it for the company. One founder I talked to was **[19:22]** spending between three and $500 a day on Google Cloud running his agent. And you can do, **[19:31]** like you're saying, a God mode CEO, you know, founder agent stack all day and you're just not **[19:39]** blocked. You know what I mean? There's just nothing blocking you along the way at all. **[19:45]** Every single one of them that I've talked to has given up on this and spun them down or kept them **[19:50]** just for themselves. Because you very quickly see for me to push this out to people, okay, **[19:57]** now it's like really a product. You know, these founders who are building agents are engineers. **[20:04]** They're engineers who are senior enough to have gone and started a company, you know, **[20:08]** some tech company. Then they have some success. They hire a sales leader. They hire their first **[20:14]** five reps. Then they hire 20 more. And, you know, maybe they forget in their daydream of rolling **[20:21]** this out. These are salespeople. These are not technical folks. They're going to interact with **[20:26]** it the way they interact with it. And you're not going to change that. How do they interact with it? **[20:33]** They reply, dude. You send somebody a slack from an agent and say, here are the deals that matter, **[20:41]** you know, such and such sales rep. And here's how to think about the meeting you have at 3 p.m. **[20:46]** Sales rep is going to go, well, what about this one or what about that one? Wait, that's not a **[20:52]** deal. Oh, wait, it is a deal. Oh, that is a deal. Oh, right. Because so and so is on vacation. So **[20:58]** I'm doing that one. Okay, right. And it's talking with the agent. So first of all, you need to be **[21:03]** able to continue the conversation. These are not just like one off web hooky, fire and forget slack **[21:09]** notifications. You need to have like a commercial chat, you know, concept that is deployed across **[21:16]** platforms. The other thing that people say when they reply, they get it in a sales rep. And by **[21:22]** the way, they're well within their rights to do this. They get something and they go, you send this **[21:27]** to me every morning. I don't want it like this. And I don't want it every day. I want it Wednesday **[21:32]** morning ahead of my pipeline meeting. And what I need is for you to like triple check that none of **[21:37]** my deals can be moved forward any more than they are because like I really want to show as much **[21:42]** progress as possible. Give me this kind of summary in the details here. And by the way, give me **[21:49]** something really smart to say in the meeting, you know, give me the talk track for my one on one **[21:54]** with why not? What like they're within their rights to tell the agent what they want, especially **[22:01]** salespeople, Anthony, where we give salespeople a lot of room to be themselves and to do things **[22:10]** their way. That's what we're paying them for. We need to kind of solve this paradox of wanting to **[22:17]** dictate their behavior and let them be themselves and hold them accountable to a result and say you **[22:26]** have the talent and intuition, instincts, rapport building, et cetera, be yourself, but be yourself **[22:32]** within like this sales process that we're telling you to follow. That's nothing in agentic coding **[22:40]** comes anywhere near solving those problems. So you can't just like grab codex or a quad code **[22:46]** and use the automation there to accomplish this goal. You can't like push it out, manage it, **[22:52]** deploy it and have your salespeople come to you and be like, I love this thing. It's not happening. ### 22:59 — The schoolyard fence: how boundaries expand what reps do **[22:59]** Yeah. And I think again, not the funnest topic, but having boundaries actually expands how far **[23:06]** people will go to. There's a famous study where they would put a fence around elementary schools **[23:11]** or not put a fence around elementary schools and the kids intuitively, you would think maybe, **[23:17]** Hey, if there's no boundary, they might just go as far out as they want to go. **[23:21]** But if there wasn't a boundary, they would actually stay really, really tight to the school **[23:25]** and really tight to the building. And they wouldn't venture out too far. If you did put **[23:28]** a boundary around the school, the kids would go all the way up until the fence and then walk around **[23:33]** and expand where they would actually go. So in this case, I think if you give some AI capabilities **[23:40]** with the right boundaries and guardrails, it'll give the sales team the flexibility they need, **[23:46]** but also give them the ability to expand and actually leverage AI in a meaningful and helpful way. **[23:53]** I had not heard that. That is incredible. And I really believe it. **[24:00]** We could take a couple examples and play with them because this is exactly what I'm talking **[24:04]** about. I was not aware of the school experiment. That is fascinating. And I totally believe it. **[24:11]** So let's say we're going to do a couple of things with agents, basic stuff like this, **[24:16]** the stuff we do with our customers. First of all, you have a CRM already. All right, great. **[24:21]** Let's eliminate all the data entry. We had a rep join our company and by Thursday or Wednesday, **[24:29]** she looked up from her laptop and said to everybody, I just don't have these CRM data **[24:35]** entry blocks on my calendar anymore. And I'm never going to in my career again. That's crazy. **[24:43]** That's amazing. And we're like, hallelujah. And we knew that, you know what I mean? But seeing **[24:50]** someone just sitting at the desk next to us, around the couches in the middle of our office, **[24:54]** say it in the look on her face, she's like, that's psychedelic. That's crazy. So that's a basic thing. **[25:03]** You probably as the RevOps super admin, founder, whatever, if you're the person in charge of the **[25:12]** sales process, if you're the RevOps person who has the real sales process in that folder, **[25:19]** and you want to push that out so that agents are updating everything on behalf of the reps **[25:26]** in HubSpot, Salesforce, whatever, you want it to be proving why it's moving an opportunity, **[25:34]** creating an opportunity, updating a, you know, med pick field, whatever it's doing. **[25:41]** And so let's say maybe you roll that out. And that has to be like a managed skill that only you **[25:47]** control. Now, what's another like really bread and butter thing that, you know, that we do with our **[25:54]** customers, but it's the type of thing everybody's going to be doing or trying to do. You know, **[26:00]** we call it meeting prep historically, it's like way different from hitting prep. This is somebody **[26:06]** grabbing you before the meeting and telling you exactly what you need to say, you know, based on **[26:13]** a deep analysis of exactly what's happening in this conversation, maybe deeply researching to **[26:19]** find similar deals with similar companies that got stuck in a similar way and how they got unlocked, **[26:25]** like, you know, just like fanned out sub agent, deep research, long horizon agent, just whatever, **[26:32]** which is like, sounds like Star Trek, but it's, it's like drinking a glass of water these days. **[26:38]** You probably want to roll that out and allow the person to have some feedback about what's **[26:45]** helpful and not as they walk into a meeting, the morning of a key meeting, whatever it is, **[26:53]** the rep has a right to say, talk to me like this, be more direct with me, you know, be less direct **[27:00]** with me, position this in a way that pumps me up, you know, tell me the stuff software engineers **[27:07]** are doing where they're writing cloud code plugins that make them do pushups, you know, **[27:11]** every couple of hours, like literally, I have not seen that. I have not seen that. **[27:18]** I might need it. If you could send me the plugins, I could probably use it. **[27:21]** Dude, and they're good. And they're like, they use the webcam to verify that you **[27:25]** actually did the pushups and they and you can't keep working until you've done them stuff like **[27:30]** that. I mean, maybe I want it this way. I want it that way. So how do you from your folder, push, ### 27:39 — The deploy button: pushing skills into personal harnesses **[27:39]** almost like a, like a Vercel or Netlify, like deploy button, push skills out, **[27:46]** but then they live within a personal harness. So you can actually take what you want to be central **[27:55]** and controlled by you, keep it that way, deploy that, take what updates you want to all of the **[28:02]** personal harnesses that are out there in the wild, that are all a result of what you've given them in **[28:08]** the feedback they've given the agents. And to be able to modify those things over time in really **[28:15]** strategic ways without sacrificing the personalization that everybody wants. Like this is going to be **[28:20]** super obvious in a year. And it's just like starting to occur to people now. That is a **[28:27]** really hard problem. We have a harness for Cloud Code and folks are like losing their minds over **[28:32]** it. You know, they just add this stuff to their, the folder they already have and Cloud Code's like, **[28:39]** I know exactly what to do. Like let's develop a plan. You know, it's got to develop a plan. **[28:45]** Give, you know, give me feedback on this plan. How do you feel about it? Okay. Here are the changes **[28:49]** we're going to make. You want to go? All right. Let's go. And it versions it, you know, does change **[28:55]** log to get stuff. I'm sure you're doing a lot of this tooling. And we're just kind of giving those **[29:00]** best practices to people in ways that like we're giving it to them through Cloud Code. So they just **[29:08]** have to understand kind of why they're getting it. And then the rest of it just all makes sense and **[29:12]** works forever. Yeah. And I think maybe going a little bit deeper on the architecture and maybe **[29:20]** specific problems that you run into, especially like, okay, when you're pushing it, which is very, **[29:26]** we mostly use Netlify. So like, okay, we're going to push an app to Netlify. We're going to push **[29:31]** this thing there, but then how is it getting customized to that particular user? So like, **[29:38]** Hey, we're going to push this thing out. We want sales directors to be able to use it a certain **[29:42]** way. We want sales reps to use in a certain way. And then the other one too, there's, there's one **[29:47]** where it's, Hey, these are personalized agents that kind of run within the parameters for you, **[29:52]** but then how do we get the team working on a kind of like open source multiplayer mode agent? And I **[30:00]** almost think of them as employees on the team. Like, Hey, this is our podcast coordinator agent **[30:06]** and anyone can operate with it. And then these are some of the parameters. How are you opening **[30:11]** up kind of those different unique scenarios? Well, you're speaking my language. First of all, **[30:18]** huge fans of Netlify, huge fans of Netlify got to hang on to that recently. **[30:23]** Claude apparently loves Vercel though. Even like, Hey, we use Netlify here. You want to push this **[30:28]** to Vercel? No, we use Netlify here. Yeah, I mean, you know, whatever. It's at a hobbyist level. You **[30:34]** know, we all use everything, I think. You know what I mean? We got stuff on every one of these **[30:38]** things. For a commercial app, I think Netlify is a great company. We invested way back in the day **[30:45]** and there's series C and I love Matt and Richard over there and everybody. Anyway, ### 30:53 — The virtual employee: rolling out agents like a new hire class **[30:53]** you're thinking about it the right way. If you ask me, the virtual employee kind of concept **[30:59]** is correct. So what is shared with the standards in software engineering, the concept of an agent **[31:06]** definition where there's kind of a job description, you know, in a code repo, your rig is going to **[31:14]** look like, you know, front end performance, you know, optimization agent or, you know, react, **[31:21]** UI designer builder, front end generalist agent, whatever. And then everybody's going to use the **[31:27]** same definition. When it comes to these virtual employees, when it comes to a seller or a CSM **[31:36]** or whoever being given an agent, they have to be able to give it feedback and have it **[31:42]** modify itself, like be coached. And so it's not that same concept of universal agent definitions. **[31:52]** You can have templates, you know, and we do, but I do actually, and we've never talked about this, **[31:59]** I do literally have an agent named Chloe who is in charge of this whole part of my life, what we're **[32:05]** doing right now, podcast stuff, you know, this type of content partnership, hanging out, doing **[32:10]** Riverside, you know, whatever it is. And that's her full-time job. So she'll slack me, she'll **[32:18]** email me, she'll give me email drafts for podcasts that she's found that are a good fit, talk to me **[32:24]** about where she found them, why she thinks they're a good fit, you know, off in the background, **[32:30]** running through based on what I've done in the past, based on how things have been received, **[32:35]** like, hey, Christopher, here's a podcast. Here's why it's interesting. You know, here's what we **[32:42]** could talk about on it that we think would be a good fit. What do you think? Okay, I have an **[32:45]** email draft ready for you, click here to send it, etc. That's the shape. Then I have a chief of staff **[32:52]** agent that is grabbing me and saying, hey, Christopher, a really interesting thing happened **[32:56]** with this deal over here. Why don't you write the CEO? Why don't you write the VP a product? **[33:02]** Because they're starting to get interested in how this thing is, like, keeping linear up to date, **[33:08]** which is something we're hearing. Like, product teams are super interested in what we're doing, **[33:13]** which I think is a good sign. A lot of fun for me as a product person. You know, that agent is saying, **[33:19]** write this, call this person, talk to this sales rep. Like, this is where you can be involved in **[33:24]** all this stuff. So that's how you accomplish it. You roll these agents out as though you were **[33:30]** introducing a new hire class, you know, to those managers. And a manager could be a support rep, **[33:40]** could be anybody at the company now has these agents that have a name, that have a job **[33:45]** description, that have skills. Those skills can run, you know, on a schedule. They can run based **[33:53]** on a meeting happening or an email coming in, whatever. We package those things together, **[33:59]** which I think is, it's one of these things. I think we're going to see a lot of apps do this. **[34:04]** I haven't seen anybody do it yet. But it's like you have a skill definition. And then boom, **[34:09]** right there, you can say, just run this every morning. Run this every time we do a meeting. **[34:14]** You know, super easy for anybody to set this stuff up. So then the workflow is, sure, you could have **[34:21]** brown bag lunches and have everybody open their laptop and like click through and build these **[34:26]** agents. It's way better for the RevOps person with their folder, with their ICP, their pipeline, **[34:35]** their big initiatives from the CEO, you know, whatever it is, to connect the day IMCP, **[34:42]** which allows it to then go do a bunch of research super fast and develop a plan. **[34:49]** Look at everybody's agents, everybody's skills, and come up with a plan of edits that it wants to **[34:55]** make. So let's say the CEO is, you know, as you know, we get very like stubborn hell bent on **[35:04]** something. So let's say the CEO's big thing is like we have this new product and we're not talking **[35:10]** about it, folks. Like we got to get everybody selling this new product that is like our baby. **[35:15]** It's the future of the company. And the sales reps are like, yeah, sure, you know, whatever. **[35:19]** They're already feeding their family selling the core product. Like what do they need this new **[35:23]** product for? Where would you update the automation? Here's where you start. You start in cloud code ### 35:32 — Changing the behavior of 1,000 reps as one change **[35:32]** by saying, let's review the actual day in a life of these sellers. Let's go look at the last **[35:39]** hundred meetings, hundred email threads, the last hundred ops, everything that happened on them, **[35:45]** maybe. Identify where we should have talked about this new product. Okay. Now come up with **[35:56]** insight about how to identify that moment. Okay, done. And especially by the way, dude, **[36:02]** with something like Fable. Oh, it's unbelievable. It comes back. It's just like, this is what you **[36:09]** need to do as a company. Like great. It's nuts. And just, just a quick sidebar. Also you were **[36:16]** talking about like speed to table stakes, then Kimmy coming out with a relevant model at a **[36:22]** fraction of the cost. Like I'm just thinking a couple of months from now, the capabilities that **[36:28]** you'll be able to have access to at really, really affordable levels. I, I don't know. And then **[36:34]** whatever the next frontier model is, I can't even really imagine what, what that's going to unlock. **[36:39]** That's why, I mean, we're priced pretty low for the capacity. You know, it's not, **[36:47]** it's something that as the model prices come down, you know, we'll be able to build a really **[36:54]** amazing business around right now. It's like, it's definitely on us. There's a lot of like open bar **[37:02]** of the, of the AI revolution here, Opus, Opus on us, you know, whatever kind of thing. But you know, **[37:07]** a lot of people are doing that. That's, that's totally fine. But being able to take those insights, **[37:13]** watch all those meetings, you know, come up with a plan and then review at a very granular level, **[37:19]** all of the automation that's there, and then deploy those changes and preserve the personalization **[37:27]** and preferences that everybody has. And actually like change the behavior of a group of a hundred **[37:34]** people, a thousand people, like it literally doesn't matter, Anthony, how many people there are. **[37:40]** And to just do that as like one piece of change management under the covers, it's just wild. **[37:47]** Well, and I think there's something to be said about how humans organize things in their head **[37:53]** as well. So if you're thinking about, okay, we're going to have X amount of workflows, **[37:58]** X amount of skills and capabilities, organizing them into roles and jobs, that is a container **[38:06]** for these things. I think it was just really intuitive. And the same way you would build an **[38:12]** organization, you'd start to specialize and have teams and these teams have access to certain **[38:16]** information. They have certain skill sets on these teams. I think it's the same way as you're **[38:21]** building your agentic empires that you would start to organize things in that way. And I just, **[38:28]** I think some of how things are built, if you're not using something like day, **[38:33]** it's just not as intuitive. So creating that container, creating that harness with all of **[38:37]** the security and governance and all the things that we've talked about amazing, but also just **[38:41]** the human intuition of, Oh, I need an executive assistant agent. They should be able to do these **[38:47]** things. Oh, somebody else on my team needs an executive assistant for them. Copy paste those **[38:52]** skill sets and all the things that work across everybody tailor it to how they like to manage **[38:56]** their calendar and their email and X, Y, Z. So, and then all the other roles you could think of, **[39:02]** I think there's so much value in just that aspect alone. **[39:09]** Of all of the really hard problems of trying to build a thing that is easy to buy, easy to use, **[39:18]** easy to build a company around on our side, there are a bunch of really hard problems. **[39:24]** The packaging and presentation of security, roles, perms, RBAC, IAM, whatever you want to **[39:30]** think of it as, the pricing, a lot of these early apps like cursor and so forth, **[39:39]** boy, they've, and God bless them. And they're doing a good thing. And it's no criticism of them, **[39:45]** but they've stubbed their toes with customers over and over simply because there wasn't any **[39:51]** intuitive concept for you're out of whatever. You know what I mean? And if we don't have some limit, **[40:00]** you're going to bankrupt the company. There was a founder recently who was saying on Twitter, **[40:06]** they had one query, they cost $20,000 and it's an amazing company, doing amazing things. They're **[40:15]** very successful, but they're going to have to solve that. You know what I mean? **[40:21]** Every one of these things, like how do you explain a quad code style, which is now by the **[40:28]** way, a codec style, agent, skill, instructions, ecosystem, that whole shape changed a lot over **[40:37]** the last three years. I know because I used, you know, I was like an OG, you know, cursor rules, **[40:43]** MD, you know, guy back in 20, like 2023. I mean, they're trying to do cursor rules. You know what **[40:51]** I mean? Now that's coalesced. It's intuitive for software engineers, mostly, not even entirely. **[41:00]** And now we have to explain this to sales reps. All of these problems actually are not problems **[41:08]** when you adopt this idea of a virtual team member. I give them a job description. They have a certain **[41:15]** set of things they're trained to do repeatedly. Any of these things they're trained to do, I can **[41:19]** ask them to do at a regular cadence or when a certain thing happens. Everybody, 100% of people **[41:25]** are like, that don't make sense. And we say, okay, and we have tiers of these. And the higher tiers **[41:32]** can do more work in a day and are smarter and have access to, you know, more powerful tools and **[41:38]** capabilities. People are like, yeah. Can we hang on that topic for a second? Because I think it's ### 41:48 — The coming "AI Operations" team **[41:48]** something people are not going to realize. I'm forecasting there's going to be a team that's **[41:55]** called AI operations within every company. And those are some of the problems that they're going **[42:00]** to have to manage of, okay, what type of agent is this and what type of model do we need to? Because **[42:08]** the subsidization of all of this, it'll go away at some point. So then we'll have to figure out, **[42:14]** okay, we can't just use the highest end frontier model for every single thing we do and just burn **[42:18]** through usage credits. We're going to have to figure out, oh, this is a simple agent. It doesn't **[42:23]** need more than this level of model. And we can manage our costs this way because there are real **[42:29]** hard costs to leveraging AI. In some cases, it's comparable to bringing on a full-time person. **[42:35]** So I know it'll not be at that point forever, but people will have to manage that. And I think **[42:42]** organizing your agents or organizing your capabilities in that way actually helps give **[42:49]** you a framework to say which model you need, what level of access you need. And that's going to be **[42:54]** a big problem too. Packaging it also in a way that is very intuitive to humans has the benefit **[43:06]** that it is also very intuitive to something like cloth. And so you could grab our product **[43:16]** and let's say you have some data laying around, you have Gong, you connect Gong, **[43:21]** all the Gong stuffs in there, Salesforce, whatever it is. You could just go in through **[43:27]** cloud code and say, talk to me about what I might do. And if you use our, we have a sort of a GTM **[43:38]** brain template repo that basically you just download or you can even give the URL, the GitHub **[43:45]** public URL to cloud and you don't even need to download it. I discovered this the other day that **[43:50]** it can just read the online stuff and it's like, yeah, yeah, yeah, I totally get it. Yeah, yeah, **[43:54]** what did you want to do? Well, you tell me, I want to see what's going on in here and it'll **[44:00]** read the instructions of basically how to think about DAI and how you might think about this. This **[44:06]** is like tens of thousands of dollars of onboarding. You know what I mean? That's just in there and **[44:12]** cloud can go, well, let me look around. You can just run a slash start command and it'll say, **[44:17]** yeah, yeah, let me go figure out what's going on. And it reads everything and watches all the **[44:22]** meetings and all this stuff comes back and it goes, okay, this makes a ton of sense. All right. **[44:27]** So for your SMB East Coast team, how about we eliminate all of the CRM data entry and **[44:37]** try to take the things that the more senior people are doing, like so and so and so and so, **[44:42]** and get the more junior people doing that stuff too. And maybe also let's do some prospecting in **[44:50]** the West Coast because those are newer territories, we don't have enough leads there. **[44:54]** And you're sitting there and you're like, well, that sounds great. What would that cost? And it **[44:58]** can tell you. And the way the whole thing is designed is not to be extractive. So if you don't **[45:07]** need a certain tier or access, you'll never hear about it. It's only at the moment that you want **[45:16]** to do something. So if you're on like our lowest tier and you never try to like manage opportunities **[45:22]** in a pipeline, you know, you never try to like create internal documentation, etc. You're just **[45:28]** like doing meetings and you're getting your notes, you're managing your to do's and all this kind of **[45:31]** stuff. You're good. You're never going to hear about it. The minute you try to do something, **[45:36]** let's say you get promoted, you know, and now you're supposed to be doing, you know, have **[45:41]** oversight and you're responsible for the hygiene of this part of the system, your agent will tell **[45:47]** you that and will actually tell your boss, which is pretty cool and say, hey, listen, Christopher is **[45:52]** in here. And, you know, his title just changed. And he's trying to do X, Y, and Z, which like, **[45:58]** I think you want him doing. And so you can click here to approve this. And I will be able to do X, **[46:04]** Y, and Z. So anyway, that like the ergonomics in the dynamism of as a vendor being able to deliver **[46:12]** the value, like just in time, we would say, not trying to oversell a package or undersell something **[46:19]** and like be too cute about it. Because here's the thing, these models are going to tell you what you **[46:24]** should be using. And they should just be able to tell you exactly how to think about it, keep it **[46:28]** super clean top to bottom. So there are no surprises and you're just good. Where do you think ### 46:37 — Where GTM is going: OTEs double, sellers become marketers **[46:37]** this is all going in the next couple months? I won't even ask you to forecast out more than a year, **[46:43]** but what are we going to see coming soon? Listen, I think that we will see some jobs **[46:54]** get really affected in a negative way. I really think we do. I don't think it's the end of the **[46:59]** world. I do think particularly the very largest enterprises are looking for a lot of efficiency. **[47:07]** And there are certain roles where they're going to get hit hard, and that's going to suck. And **[47:10]** we're all going to have friends that get hit. My recommendation to everybody is like, just get into **[47:16]** this AI stuff. And if you have a period, and I've had lots of friends have periods of not being, **[47:22]** you know, in a role at a given time, use that time to just train, train, train, take a hobby, **[47:29]** take an interest, whatever it is, your softball team, your church, anything, and use these frontier **[47:35]** models to get super deep. And that is the way to feed your family long term, 100%. Like I think **[47:42]** we're all going to be okay, but we're going to have to rescale a bit. That said, you know, and go to **[47:48]** market, we will see reps basically doing a lot more business straight up. And Mark Roberge is **[47:58]** out there, you know, I'm sure you know Mark well. He's doing a very clever thing talking about the **[48:05]** number of active opportunities being worked by a rep going up by, you know, could be 30% at a **[48:11]** company could be 60% at another company. And that, you know, you don't have to squint too, **[48:17]** too much to see that that turns into higher OTEs, you know, I think OTEs for sales reps are going **[48:23]** to double. My finance plan has OTE doubling over 24 months here. So, you know, these reps, **[48:32]** they come in, they make 200k, they're going to make 400k, even at the same level, just through **[48:38]** the efficiency through not having to waste their time doing stuff, you know, they don't need to be **[48:41]** doing these inside sales reps, if they're doing two quality calls a day, you know, 2.5 or something **[48:46]** like that. They're happy to do four. They're happy to do four. And the company is happy to pay them **[48:53]** the OTE to close twice as much business. Sales management, you know, I'm in the territory of **[49:00]** what is starting to be consensus forecast. Sales managers will go from seven direct reports to **[49:09]** twice that. That'll be really interesting. Sales managers have a tough job. A lot of them are new **[49:15]** to management. That's one area where I think we internally want to be locking arms with them and **[49:21]** thinking what we can do to help their careers, give them lenses, tools, content, you know, **[49:26]** encouragement. I think some of those folks are, you know, going to hit a tough patch. Some of them **[49:36]** will just be given a lot more work. The ones who really embrace this stuff, I think, will have a **[49:42]** great quality of life and make a bunch more money. Beyond that, I think that we'll see roles start to **[49:48]** merge. I see a senior seller, for example, starting to do a lot more marketing as we specialize. **[50:00]** What do I want as a buyer? I want to, you know, ask a question about whatever it is that I'm **[50:06]** interested in. I want to get a great answer from Claude or Chachi B.T. or Perplexity. You know, go **[50:12]** to some citation. See a great offer or piece of content with somebody's name and face on it as **[50:20]** the author. And then when I talk to that company, I want to talk to that person. And I want to work **[50:27]** with that person the whole time I work with that company. I mean, this is like, if you're Coca-Cola, **[50:31]** this is what life is like as a buyer. We've always, in SMB mid-market scale-ups, always wanted that **[50:38]** to be the buyer experience. I think that becomes the buyer experience. We get specialized in sales **[50:44]** and we start doing a lot of the customer marketing, the case study stuff, the testimonial stuff, **[50:50]** because why not? You're a rep, you have a book of business. Now maybe you know this type of company **[50:56]** within that book of business. That's your whole job. You can make a landing page, dude. You can **[51:00]** make an ebook. You can do a webinar. That's no problem. And as a buyer, oh, it's wonderful. **[51:07]** I think those people become indispensable for any company. That's a really interesting take. **[51:13]** And I haven't heard that one yet. And it's one that, you know, as a business, we've leaned very, **[51:21]** very heavily into content. A lot of companies have, especially your alma mater HubSpot. We just had the **[51:27]** VP of media on the podcast a couple of weeks ago. And it's an unbelievably powerful tool that also **[51:36]** feeds the LLMs so well. And I think you're right. And the best sellers **[51:44]** do already have a personal brand and they bring that personal brand between companies. They start **[51:49]** to concentrate in a particular industry or on a particular type of solution. So I think the **[51:55]** evolution of sellers going into marketing makes a ton of sense. And now I think before they were **[52:02]** just held back by sheer bandwidth. Oh, well, who's going to design my webpage and who's going to edit **[52:10]** what I write and who's going to come up with ideas for my posts or whatever it is. **[52:15]** Now they can have an entire marketing agency underneath them that could power the entire process. **[52:23]** A hundred percent. I mean, just the amount and the way in which I'm in touch with our customers, **[52:32]** it's just so much more scalable than it used to be. You know, it's easier for them to write me **[52:38]** with a question they have. It's easier for me to respond to them with, you know, documentation, **[52:45]** let's say, or something, you know, it's just a wonderful interaction now. I mean, somebody that, **[52:50]** you know, I talked to in the sales process six months ago, nine months ago, I got an email from **[52:54]** from a guy over the weekend and he said, oh, we're trying to do this thing with, you know, the API. **[53:00]** And, you know, we're have a question about how to do the auth. And I go to just take this GitHub link **[53:08]** and just tell cloud code about it. And he writes me back. He's like, okay, wow. **[53:14]** But, you know, and so he and I just get to have this wonderful touchpoint. It's like, by the way, **[53:19]** how you been? You know what I mean? We have like cloud code and codecs and all this stuff. Like, **[53:25]** I ran into that question one time. I created, you know, resources, example apps, all of this kind **[53:33]** of stuff. And all I ever have to do if somebody asked me a similar question is go, hey, just tell **[53:38]** cloud code to go look here. And any question it could possibly have on the topic, it's going to **[53:43]** find reference examples, solve the thing. And now this other human and I, we get to talk about our **[53:48]** kids. We get to talk about that trip he took to Europe. You know what I mean? Work is getting **[53:53]** really fun, I will say. It's getting really fun that we're focused on the creative side 100% of **[54:00]** the time. Yeah, I agree. And I think it's tough unless you've really unlocked certain things. But **[54:07]** I think about running lean scale as an agency, some of the things that I would have to do, **[54:13]** I'd have to have tons of one on ones where they're just regurgitating information back to me. **[54:19]** We would spend hours handwriting documentation after a meeting. So we've become I would, **[54:29]** it's fair to say we're at least twice as productive now. And all of those things that used to take so **[54:35]** much time, or if I wanted to model out some type of thing, I would open up Google sheets and I'd **[54:40]** manually type in everything or download data and try to like dedupe it, restructure it for myself. **[54:46]** Now I can just connect the MCPs and get things running. I agree. I get to do more fun work. I **[54:53]** can do more podcasts. I can do, I could create more content. I can spend time ideating with the team. **[55:02]** Also another thing that's just opened up, like let's just hop on a meeting. It's going to get **[55:05]** recorded. Even you mentioned this earlier in the conversation, when we have off-sites, **[55:11]** we can spend more of it just doing fun stuff together and just getting to know each other **[55:15]** as humans. And then when we're having meetings, just throw a recorder in the middle and then **[55:19]** just talk. We don't have to worry about, no one has to describe it, no one has to take notes, **[55:22]** no one has to be worried about anything. Just let the ideas free flow and then we'll get it **[55:26]** all structured later. It's made my work day way more enjoyable. I think you bring up a very **[55:35]** very important image of the before and after that had not occurred to me. **[55:44]** If one were to listen to the hallway conversations or see the internal slacks and email threads, **[55:55]** here, I will say here because it's what I know because I work here and we're very far down this **[56:00]** path ourselves. One would very quickly notice the density of the questions, answers, the **[56:13]** very very high value information exchange. I don't have to take 90 minutes with the head of sales **[56:24]** to see a PowerPoint presentation of here's how the reps are doing and here's the thing **[56:29]** and here's the thing. I know all of it. I know all of it, Anthony. Somebody puts a meeting on **[56:37]** the calendar. When you and I first met, Sam didn't have to grab me and be like, "Let me explain who **[56:43]** Anthony is." I know who you are. I know your relationship with him and this other person and **[56:50]** this other person. I've gone and checked this stuff out. I know all of that. I don't need to **[56:57]** be given a PowerPoint presentation. I'm pointed to one very key question about here's a rep who **[57:08]** had a deal and that deal didn't go our way. Now, you could make the case that he ought to have **[57:16]** positioned it this way as opposed to that way. I go, "Oh, well, that's interesting to me as a CEO. **[57:22]** That's extremely interesting." I dive in. I'm watching clips. I learn everything about this **[57:29]** one deal. I spend maybe 10 minutes. I know everything about this deal in 10 minutes **[57:35]** and then I draft up an email with my agent to the head of sales and I say, "Should we talk about **[57:41]** this? Is there a debate to be had here? Is this a simple matter of coaching? What's going on with **[57:46]** this one?" I get back minutes later an equally dense and high-value explanation of the entire **[57:56]** topic of whether to position it this way or that way. What's hard about it? This type of prospect, **[58:02]** our overall go-to-market, everything, dude, everything. It's just like unbelievably good and **[58:07]** I read it and I'm like, "Ha." Then I go back and with that new lens from a colleague, I look at it **[58:14]** again and I write her back and I go, "You're completely correct." I won't say you're absolutely **[58:22]** right, but I'm like, "I completely agree with you. Thank you." That whole interaction was **[58:31]** so much less stressful and so much higher density for everybody involved than it ever would have **[58:38]** been. We can spec out entire modules of software we want to build just sitting around the couches **[58:46]** with marketing, engineering, sales, just talking about it. We've done this entity three or four **[58:52]** hours and then you take the recording of that and you start to shape that into the very detailed **[59:00]** picture of what you want to build. Once you have that, coding is like ironing that on the code base, **[59:06]** like a sticker. You know what I mean? That part's easy, free. I'm enjoying work and people here **[59:14]** often say, like our head of product is famous for saying, "I'm never working again without **[59:20]** working this way." No, you get to do the most enjoyable things. You get to connect with people, **[59:26]** share ideas, you get to talk, you have conversation, you can make it enjoyable, grab coffee together **[59:30]** and just make sure it's recorded and then dump everything in and you get higher value outputs. ### 59:37 — Mathematicians with calculators: the atrophy question **[59:37]** One question I have, flip side, I'm not sure where I stand on this one. We are at a unique time where **[59:45]** we've done all of the work in manual mode, so we have an appreciation for the depth that things **[59:52]** should go and then we're equipped and powered with automation and things that can just **[59:57]** supercharge us. I would compare it to we are mathematicians that just got handed calculators. **[1:00:05]** Now, fast forward 10 years or so. Are there any concerns around **[1:00:12]** atrophy and the ability to leverage these models if you don't have some baseline of skill **[1:00:21]** as the foundation in the first place? I think it is a genuinely brilliant question **[1:00:29]** and extremely relevant. This is outside of the one year time frame, I would say really the question. **[1:00:39]** So in software engineering, we've had this kind of magical level of living available to us now for **[1:00:47]** a little while. It has not caused people to work less. It has not caused people to be **[1:00:58]** less stressed. We are not getting twice as much done in 60% of the time and then going and touching **[1:01:08]** grass and throwing football with her daughter in the backyard or whatever. I mean, I'm starting to, **[1:01:15]** and four or five months ago, I think I started forcing myself to. The article started hitting **[1:01:21]** around brain fry and that becomes very real. When you get exposed to this stuff and you learn these **[1:01:32]** best practices and ways of multi-threading and kind of hypercharging your work, the tendency is to do **[1:01:40]** five times as much work in 140% of the time. You know what I mean? Yep. I do think we will see a **[1:01:50]** lot of people from this era who went really deep early just run because I don't know the neuroscience **[1:02:01]** behind it, but boy, I mean, you just end up just kind of catatonic. If you really are doing **[1:02:08]** everything, because think about it, it's like I'm doing landing pages, I'm doing this, I'm going to **[1:02:11]** make another offer and then let me go over here and do a feature. Oh, why would I ask engineering **[1:02:16]** for this feature when I can write this back with quad code? If I have the spec, then the automated **[1:02:22]** sandbox thing is going to pick it up and just code it and then somebody's going to do a pull request **[1:02:27]** and this thing's live. There's all this stuff to be checking in on and you're leaving Thanksgiving **[1:02:36]** dinner halfway through to check on your quad code processes and keep them going and unblock them. **[1:02:41]** And it's like sneaking away to unblock stuff on your phone. I think we'll get tired of it. **[1:02:48]** I think we will. And I don't think it's a good thing. And it's something I've tried to really **[1:02:55]** change in my own behavior. So I don't know. My hope and what I say internally is for people to **[1:03:02]** be twice as productive in 80% of the time, I want to see a salesperson work 80% of the hours they **[1:03:09]** work today have all of that time be what they consider their value add and make twice as much **[1:03:18]** fun. Great. That's our kind of North Star on this thing. I think a lot of roles will get there. **[1:03:26]** Some roles will go away and we don't quite know which is which. So it's definitely something to **[1:03:33]** stay really close to and something that I'm a student of and I don't know where we're going **[1:03:38]** to land, but I think we're going to see a lot of goat farmers, bro. Well, I'm also thinking one **[1:03:45]** quick parallel. Like if you go back 100 years ago, industrial revolution, really picking up steam, **[1:03:53]** something that wasn't around was a fitness center and gym on every single corner of the world **[1:04:03]** because you just did the fitness throughout your work. Like you're, you're lifting things. You're **[1:04:08]** moving things. And then we go into the knowledge era where you're actively like, I'm doing **[1:04:16]** relatively complicated math problems throughout the day. I'm structuring my thoughts by manually **[1:04:21]** typing them out throughout the day. I'm reading, you know, dense material or going through research **[1:04:26]** process and I'm using my brain quite a bit. I'm wondering if there's a world where you fast **[1:04:31]** forward, we're going to need to do some math by hand and read books on paper just to keep our **[1:04:38]** brains fresh and productive for leveraging the AI we have access to. **[1:04:44]** I mean, listen, I think I carve out time to understand as well as I can what this all means **[1:04:56]** for creative pursuits that have nothing to do with our jobs, you know, our day jobs. **[1:05:07]** I, I seek to understand this in depth. My instinct is that the upshot of all of this, **[1:05:16]** in the good case, is we get to do art. We get to play music. We get to sit around a campfire, **[1:05:25]** you know, lazily for long periods of time and tell stories to each other. I mean, **[1:05:29]** this is a very optimistic rose-colored glasses, you know, it's been said there's a band I really **[1:05:38]** like. They describe it as nostalgia for the future. But so sue me. I mean, what else am **[1:05:44]** I going to do? I'm not going to be a doomer about the whole thing. So what does it mean for us to, **[1:05:50]** can the AI help us, you know, learn an instrument? It can help a 13 year old, you know, write a song **[1:05:57]** that they have in their head. We all know that we have Suno, you know, that's great. **[1:06:02]** But can it help us learn to play cello and in our 40s? Absolutely it can. Can, can it help me **[1:06:11]** improve my tennis serve so that I can be good enough to get, you know, calls from people I **[1:06:18]** meet at tennis clinic? It, you know, it absolutely can. Then what happens? Well, then 20 years from **[1:06:25]** now in my 60s, I'm playing tennis and I have a social network and, you know, the stats say **[1:06:30]** I'll live 10 years longer. So I'm spending a lot of time outside of work thinking about what are **[1:06:36]** the proxy, you know, impacts that this could have that are less obvious, that improve quality of **[1:06:43]** life and health outcomes and all the rest of that. Not trying to, you know, offer a product and build **[1:06:50]** a business and raise money and, you know, grow revenue, but just trying to see what's going to **[1:06:55]** happen. You know, what does it mean for education? Do my kids go to college? Like is college useful? **[1:07:00]** I have no idea. Is it twice as useful? Because now you actually more than ever need a four year **[1:07:07]** sandbox as your prefrontal cortex develops to just meet people, fellowship and experiment **[1:07:14]** intellectually. You know, I don't know which direction it goes, but I spent a lot of time **[1:07:19]** thinking about it. And I think that whole area is very interesting. Yeah. And I'm not a, **[1:07:28]** I'm not a doomer about it either. I just think there are going to be some interesting things that **[1:07:32]** we will end up needing to solve for. And that's just one that I think about a lot as like, okay, **[1:07:38]** I'm used to using my brain a lot. Now I don't have to use it as much, but I'm right at the point where **[1:07:44]** I do have a well working brain, one that's been used to doing difficult things. And now I can **[1:07:51]** use that plus the superpower, but I don't know if like, I mean, you've probably seen in a meeting, **[1:07:56]** somebody show up with some AI built stuff, but they don't really know what they're talking about. **[1:08:01]** And you're like, Oh, you just, you just prompted a few things. But if you don't actually know **[1:08:08]** the subject, then you don't have taste for what good looks like in this area. So I just think **[1:08:15]** we'll have to carve out some intentionality to work our brains. And I think I'm with you as well. **[1:08:21]** The most enjoyable things are spending time with your family, spending time with friends, **[1:08:25]** being active, being out in nature. And if we can carve out more time to do that, I'm all for it. ### 1:08:31 — The 12th draft: writing the blank page yourself **[1:08:31]** I'll tell you one, one thing I found myself doing is, uh, the other day I was working on a document **[1:08:36]** and, you know, a memo, like they used to say back in the day, I did, I think 11 versions of it, **[1:08:44]** molding it, shaping it with Claude, come at it this way, take into account that stuff. **[1:08:51]** And then the 12th version, I opened up Google docs, a blank document, and I looked at nothing. **[1:08:59]** And I just sat and I just wrote. And so the first 11 versions and all of the AI and the web searches **[1:09:07]** and the, you know, mashing concepts up and finding patterns and everything, that was, it turns out all **[1:09:14]** to tease my imagination for a couple of weeks. After that, I'm like, I need to do this top to **[1:09:23]** bottom myself. And it felt really good. And I felt ready for the blank page. Um, so that kind **[1:09:31]** of stuff I think is fun too. I think it will surprise ourselves. I think so too. **[1:09:37]** Christopher, this has been awesome. Thank you so much for spending so much time going through **[1:09:42]** the world of AI as of mid July, 2026. We all know it's going to change and I appreciate **[1:09:50]** any bit you're willing to forecast on it. But I think to recap some of the major themes that **[1:09:56]** we talked about, everyone has been explosive with their creativity and things are building **[1:10:03]** and building things that are really moving the needle, adding productivity and making the **[1:10:07]** difference. The challenge is when you want to share that great work across an entire organization **[1:10:14]** and when you want to enable multiplayer mode on specific capabilities. And I think what you're **[1:10:20]** doing at day AI is really unlocking that for companies at scale and very, very exciting. **[1:10:27]** And I'm also happy that you now share the positive view and outlook on the future, **[1:10:32]** maybe not the near future. We'll have some bumps as every change brings, but I think all in we're **[1:10:40]** building this for ourselves and expecting the benefit for us and our families. So, Christopher, **[1:10:46]** thank you so much. Thanks for being on the show and can't wait to see what you build at day AI **[1:10:53]** next. Can't wait to see what we do together. And until next time, yeah, this is a lot of fun. Thank **[1:11:00]** you so much for having me on. --- _LeanScale Knowledge Hub. Free to quote and cite with attribution to The LeanScale Podcast (https://www.leanscale.team)._