The LeanScale Podcast · Episode 97

Stop Making Decisions, Start Making Bets

Yishi Zuo of Tavus on poker, expected value, and why the right go-to-market call can still lose

Yishi Zuo · Head of Go-to-Market Strategy · Tavus Hosted by Anthony Enrico
Published Updated 00:42:24 43 min read 8591 words
GTM Strategy AI in GTM Revenue Operations Demand Generation
▸ Listen on Spotify Watch on YouTube Apple Podcasts
SummaryTakeawaysFrameworksQuotesAdviceAIEntitiesMetricsFAQTimelineTranscript
Executive Summary

The one-paragraph brief, extended

Why this conversation matters — and who should spend the hour.

Most operators believe they are making decisions. Yishi Zuo's argument is that they are making bets and have not noticed. Now Head of Go-to-Market Strategy at Tavus, the conversational video AI company, Yishi arrived at go-to-market by an unusually indirect route: Goldman Sachs investment banking, three years at a hedge fund, an MBA at MIT Sloan where he co-founded the expert network DeepBench to seven figures and an exit, four years running a startup studio, and a stint leading finance at a YC-backed healthcare AI company. He is also a serious poker player, and this conversation with Anthony Enrico is about what that game teaches an operator that finance alone does not.

The core transfer is expected value. Yishi works a worked example on air: a hundred-thousand-dollar investment with a fifty percent chance of returning five hundred thousand and a fifty percent chance of returning nothing carries an expected value of two hundred and fifty thousand. Then he immediately supplies the two caveats that break the arithmetic. The first is opportunity cost — the same hundred thousand may have four other homes with higher expected values. The second is ruin: if it is the last hundred thousand you have, a fifty percent chance of losing it is not a bet you can take, and a ninety-nine percent chance of returning a hundred and twenty thousand is the better call despite the lower expected value. The point is not to compute precisely but to compute at all.

From there the conversation goes to process versus outcome, which Yishi treats as fundamental to poker, investing and business alike. The right bet can lose, and being at peace with that is the hard part. But the discipline has a limit, and he draws it with an example: hold pocket aces against seven-deuce, go all in repeatedly, and you should win about ninety percent of the time. Lose five in a row and the probability is under a hundredth of a percent — at which point the correct conclusion is not bad luck but that the deck is rigged. The lesson for operators is that separating process from outcome is not permission to ignore outcomes forever; a long enough losing run is information about the process.

Anthony presses on the decision-versus-bet distinction and they land it together: a decision has broadly known outcomes — eggs or yoghurt, water or coffee — while a bet has real variables outside your control. Most business calls are implicit bets that nobody stopped to price. Yishi is careful that this does not mean modelling everything, because the mental burden is real and some fundamentals should become muscle memory rather than decisions. He also notes the option people forget: making no decision is a decision, and it has consequences, because a competitor takes the market you declined to enter.

The practical demonstration is how he runs Tavus. Conferences and outbound agencies are both structured as explicit bets: cost of tickets, travel and a rough salary estimate for the time, against a target number of strong leads at a target ACV. The model takes about ten minutes and he is candid it may be wildly off — the value is the forcing function and the opportunity-cost comparison it enables, not precision. He closes on where conversational video AI is going: digital twins of celebrities and CEOs, companionship in senior care, and B2B learning and development as the entry use case, with customers reporting two-to-five-times conversion increases in sales, sixty percent higher engagement and three-hundred percent faster ramp in L&D, and twenty-three percent more completed AI interviews than audio alone.

Key Takeaways

15 things worth stealing

The load-bearing ideas, each with the business implication and who should care.

01

Nearly every go-to-market decision is an unpriced bet

A decision has broadly known outcomes; a bet has variables outside your control that can derail the plan. Yishi's view, reached live in the conversation, is that most decisions are implicit bets of some sort, if not explicit — and that most operators never stop to notice which one they are making.

Why it matters: Before a significant call, ask explicitly whether the outcome is known or probabilistic. If it is probabilistic, it deserves an expected-value sketch rather than a discussion.

FoundersRevenue ExecutivesRevOps Leaders
02

Expected value is the most transferable framework from poker to business

A $100k investment with a 50% chance of returning $500k and a 50% chance of returning nothing has an expected value of $250k. Yishi works the arithmetic on air precisely because it is simple enough that there is no excuse for skipping it.

Why it matters: Attach an expected value to every material go-to-market investment. The number does not need to be right to be useful; it needs to exist so it can be compared.

FoundersRevOps LeadersMarketing Leaders
03

Two caveats break the expected-value math: opportunity cost and ruin

The same $100k may have four other homes with higher expected values, which makes the headline number meaningless in isolation. And if it is the last $100k you have, a 50% chance of losing it is unacceptable — a 99% chance of returning $120k is the better bet despite the lower expected value.

Why it matters: Never evaluate a bet's expected value alone. Compare it against the alternatives for the same resource, and check whether losing it is survivable.

FoundersRevenue Executives
04

Separate process from outcome — but not indefinitely

The right bet can still lose, and Yishi says being at peace with that is where people struggle most. The limit is a losing run long enough to be improbable: pocket aces against seven-deuce should win about 90% of the time, and losing five straight has a probability under 0.01%. At that point the deck is rigged, not unlucky.

Why it matters: Defend good process against a single bad outcome, and investigate the process when the losing run exceeds what variance can explain. Both halves are required; only holding the first is how a broken process survives.

FoundersRevenue ExecutivesSales Leaders
05

Fundamentals beat fancy, and reading tells is not a beginner's game

Yishi's example is how you hold your cards when you look at them — basic to the point of sounding trivial, and something beginners get wrong. Master the first layers and you are better than ninety-five percent of beginners. Tells and reading people are genuinely advanced and do not matter until then.

Why it matters: Audit whether a team is reaching for sophistication before competence. Time spent on subtle signals while the basics are unreliable is time taken from the thing that would actually move performance.

Sales LeadersRevOps LeadersFounders
06

Know what game you are playing before optimising it

Yishi starts with goals, not tactics: some people play poker to make money, others for fun, and some are at a table with prospects. In the last case losing a few hundred dollars to build relationships that lead to hundreds of thousands in ARR is a win, not a loss.

Why it matters: Define what a motion is actually for before measuring it. A channel judged against the wrong objective reads as failure while doing exactly what it was funded to do.

FoundersMarketing LeadersSales Leaders
07

A conference or an agency can be modelled in about ten minutes

Cost of tickets, travel and a rough salary estimate for the time invested, against a target number of strong leads at a target ACV. Yishi says three or four significant figures is close enough, and the whole model takes roughly ten minutes.

Why it matters: The barrier to pricing a go-to-market bet is not analytical difficulty, it is the habit. Ten minutes and an admittedly rough number beats an unexamined spend.

Marketing LeadersRevOps LeadersFounders
08

The model's value is comparison, not precision

Yishi is explicit the estimate may be wildly off. Anthony extends it: even an imprecise model lets you compare a week at a conference against the cold calls or campaign that week could have funded instead. Magnitude can be wrong while the ranking stays useful.

Why it matters: Do not abandon the estimate because it cannot be accurate. Its job is choosing between two uses of the same resource, which survives considerable error.

FoundersRevOps LeadersMarketing Leaders
09

Attribution being hard is not an excuse for having no number

Yishi concedes marketing attribution is fundamentally harder than sales, then refuses the conclusion: a $100–150k rebrand should still carry an expected value, even one as rough as raising the probability of the next round by ten or twenty percent. Thirty seconds of thinking is enough for some of it.

Why it matters: Require a stated assumption for unattributable spend rather than exempting it. Reasonable assumptions can be wrong and revised; an absent one cannot.

Marketing LeadersFounders
10

Making no decision is a decision with consequences

Declining to enter a new market is itself a bet, and it carries the outcome that a competitor enters it instead. Inaction reads as caution while carrying real downside.

Why it matters: Price the cost of not acting alongside the options on the table, so a default of inaction has to win on merit rather than by avoiding scrutiny.

FoundersRevenue Executives
11

Not every decision should get the analytical treatment

Yishi guards against the obvious over-correction. Mental burden is real, and good fundamentals should become muscle memory rather than repeated decisions — the way a poker player stops deciding how to look at their cards. Some strategic questions are also genuinely unknowable, and at some point you commit on the basis that you know enough.

Why it matters: Reserve explicit bet-pricing for material, reversible-at-a-cost calls. Turning routine work into modelling exhausts the capacity you need for the decisions that matter.

FoundersRevOps Leaders
12

Postmortems get harder when politics and clients are in the room

Yishi notes that separating process from outcome is clean in theory and complicated in practice: with clients, and inside companies with internal politics, how a failure is communicated has to be handled delicately. Knowing what to do and actually doing it are different problems.

Why it matters: Design the retro so a correct-but-losing bet can be reported without someone absorbing blame, or the honest post-mortem simply will not happen.

Revenue ExecutivesFounders
13

Pattern recognition is the compounding return on reps

After enough hours, the better player notices what deviates from the norm and interrogates it — is the sudden alertness a strong hand, or a coffee? Yishi maps this to a thousand or ten thousand sales calls, and to why agencies exist: they have seen far more of a specific problem, even without the domain depth of the company itself.

Why it matters: Value accumulated reps on a narrow problem as a distinct asset from domain knowledge. Anthony's counterpart observation is that ten thousand hours in still adds new patterns.

Sales LeadersRevOps LeadersFounders
14

Conversational video AI is showing measurable ROI across use cases

Yishi reports customers seeing two-to-five-times increases in sales conversion rates, sixty percent increases in engagement and three hundred percent faster ramp in learning and development, fifty-four percent higher confidence scores, and twenty-three percent more completed interviews than audio alone.

Why it matters: Treat these as vendor-reported customer outcomes rather than independent benchmarks, but note the pattern: the returns cluster in training and enablement, not only in customer-facing sales.

Revenue ExecutivesSales LeadersCustomer Success
15

The scaffolding, not the avatar, is the adoption barrier

Tavus's customers today are largely AI-native companies building on the API. An AI human is only useful once it is connected to a knowledge base and the surrounding systems, and that scaffolding needs AI engineers most companies do not have in house.

Why it matters: The constraint on deploying conversational AI internally is integration capacity, not model quality. Budget for the connective work or buy it, rather than assuming the avatar is the project.

Revenue ExecutivesRevOps LeadersCustomer Success
Frameworks Discussed

6 named models

Every framework Jimmy names, defined and time-stamped.

Expected Value for Go-to-Market Bets

14:02

Multiply each possible outcome by its probability and sum them: a $100k bet with a 50% chance of $500k and a 50% chance of zero has an expected value of $250k. Then check it against opportunity cost and against whether losing is survivable.

Yishi calls this the most applicable framework carried from poker into business logic. The two caveats are load-bearing — the same capital may have better homes, and a bet you cannot afford to lose is not made acceptable by a favourable expected value.

Process Versus Outcome

16:34

Judge a decision by the quality of the reasoning available at the time rather than by how it turned out — while treating an improbably long losing run as evidence the process itself is broken.

Fundamental to probability, investing, trading and poker. Yishi's pocket-aces example sets the boundary: losing five all-ins as a ninety percent favourite has a probability under 0.01%, so the right conclusion is that the deck is rigged, not that variance is being unkind.

A Decision Versus a Bet

19:28

A decision has broadly known outcomes — eggs or yoghurt, water or coffee. A bet has material variables outside your control. Most business calls are bets that were never priced as such.

Reached jointly in conversation: Yishi's position is that most decisions are implicit bets, and Anthony's addition is that operators sit in meetings making bets without pausing to evaluate the expected value of the one they are making.

Fundamentals Before Tells

11:35

Master the basic disciplines first — they make you better than ninety-five percent of beginners — and only then work on reading signals, which is genuinely advanced and does not matter until the fundamentals are automatic.

Yishi's example is holding your cards so nobody sees them: trivial-sounding, commonly done badly, and a matter of discipline rather than talent. Working on tells before that is a distraction from the thing that would actually improve results.

The Ten-Minute Bet Model

30:02

Cost of the investment plus a rough estimate of the time it consumes, against a target number of leads at a target ACV. Three or four significant figures is close enough, and the model takes about ten minutes.

How Yishi structures conferences and outbound agency spend at Tavus. He is explicit it may be wildly off; the value is that the bet gets priced at all and can then be compared against the alternative use of the same time and money.

Making No Decision Is a Decision

24:21

Declining to act is itself a bet, carrying the consequence that a competitor takes the opportunity you passed on.

Raised by Anthony as the option operators forget to put on the table. It reframes caution as a position with a cost rather than as the absence of one.

Best Quotes

22 lines worth clipping

Pulled verbatim. Copy or share any of them.

“You have to separate the process from the outcome.”
Yishi Zuo 01:08
“Let's say you think you're following the right process and you get bad outcomes every single time, and maybe you need to reevaluate your process.”
Yishi Zuo 01:10
“I think that's probably the most applicable like framework in from poker to a lot of like business logic.”
Yishi Zuo 01:42
“A lot of people tend to see the game of go to market as a little bit more subjective, a little bit more art than science. But the more time I've spent in go to market, the more it feels like a science.”
Anthony Enrico 07:35
“I try to be rational about like how I place my bets.”
Yishi Zuo 07:29
“I'd argue it's worth losing a few hundred bucks to build relationships that could lead to a few hundred thousand.”
Yishi Zuo 11:25
“If you work on the first layers of fundamentals, like you'll be better than like ninety five percent of beginners. Like don't even worry about like tells and reading people.”
Yishi Zuo 12:33
“I think being exceptionally well at the basics is how to be exceptional.”
Anthony Enrico 12:53
“There's a lot of factors that go into play. There's like the opportunity cost of like what you spend your money on.”
Yishi Zuo 15:00
“It was the right bet to make. But sometimes they don't pan out and that's OK.”
Yishi Zuo 16:11
“People have a really, really tough time doing like being at peace with making the right bet, but the bet not panning out.”
Yishi Zuo 16:16
“I think most decisions are implicit bets of some sort, if not explicit.”
Yishi Zuo 20:21
“A lot of times people are in meetings, having calls, making decisions on things and not realizing that they're actually making a bet.”
Anthony Enrico 21:16
“Attribution is like just fundamentally hard. We should still have like a model.”
Yishi Zuo 18:49
“It doesn't you don't have to spend like five hours on you could spend like five minutes or like 30 seconds.”
Yishi Zuo 19:02
“Making no decision is a decision too. So doing nothing is a route to take, and that has consequences as well.”
Anthony Enrico 24:21
“You have to make a call and do something you have to make a play, you have to double down fold, whatever. And I think in the long run, you should start to rack up the wins, but you have to be okay making it through those losses sometimes.”
Anthony Enrico 24:32
“The better you are, the more things you notice that like deviate from the norm and very quickly like hone in on that.”
Yishi Zuo 27:37
“You just really need to account for that and then go to this conference and what's the expected outcome.”
Yishi Zuo 30:53
“It literally takes probably 10 minutes and it may not be accurate could be wildly off hopefully in the good direction but at least like we tried.”
Yishi Zuo 31:08
“Maybe the magnitude could be off but at least it can help you choose between two things.”
Anthony Enrico 31:59
“It's just a muscle like you'll get better at making these calculations questioning assumptions and your assumptions will just get better too.”
Yishi Zuo 32:25
Practical Advice

What should you actually do?

The playbook, split by the seat you sit in.

Founders

  • Ask of any significant call whether the outcome is known or probabilistic. If it is probabilistic, you are making a bet and it should be priced as one.
  • Check every expected-value number against two things the arithmetic hides: what else the same resource could buy, and whether losing it is survivable.
  • Put the option of doing nothing on the table explicitly, with its cost — a market you decline to enter is one a competitor takes.
  • Defend a correct bet that lost, and investigate the process when the losing run is longer than variance can explain. Doing only the first protects a broken process.
  • Reserve this treatment for material decisions. Modelling everything exhausts the judgement you need for the calls that matter.

RevOps Leaders

  • Build the ten-minute model as a standard: investment plus loaded time cost, against target leads at a target ACV. Three significant figures is close enough.
  • Present the estimate as a comparison between two uses of the same resource rather than as a forecast — the ranking survives error that the magnitude does not.
  • Require a stated assumption for unattributable spend such as a rebrand, even a crude one like shifting the odds of the next raise by ten percent.
  • Let proven fundamentals become process rather than repeated decisions, and spend the analytical budget on the genuinely uncertain calls.

Marketing Leaders

  • Model conferences before committing: tickets, travel, and the salary cost of the people attending, against the leads and ACV you expect.
  • Do not let attribution difficulty become a reason to carry no number at all — a rough assumption can be revised, an absent one cannot.
  • Define what a channel is for before judging it. A relationship-building motion measured on direct pipeline will read as failure while working.

Sales Leaders

  • Fix the fundamentals before coaching subtlety. Reading a buyer well is an advanced skill that does not compensate for unreliable basics.
  • Treat accumulated reps on a narrow problem as a real asset — the pattern recognition from thousands of calls is what separates good from great.
  • Run post-mortems that can distinguish a bad bet from a bad outcome, and structure them so the honest version is safe to say out loud.
AI Takeaways

How AI actually changes GTM

LeanScale's signature read on the AI-in-GTM question this episode wrestles with.

The thesis

The interesting AI content here is not the avatar but the adoption constraint around it. Tavus's customers today are largely AI-native companies building on the API, because an AI human is useless until it is wired to a knowledge base and the surrounding systems — and that scaffolding needs engineering capacity most companies lack. The bottleneck is integration, not model quality.

Agent & automation ideas

  • A discovery agent for consulting engagements that clients can meet at any hour, removing hours of scheduled human discovery — Anthony's own idea in the episode.
  • An expected-value assistant that prices a proposed conference or agency spend from cost, loaded time and target ACV before the commitment is made.
  • A decision journal that records the assumptions behind each go-to-market bet so post-mortems can separate reasoning from outcome.
  • A variance monitor that flags when a motion's losing run has exceeded what its assumed win rate can explain — the operational version of the pocket-aces test.
  • Internal L&D avatars connected to the company knowledge base as a lower-risk first deployment than customer-facing agents.
Operations Takeaways

By function

The same conversation, filtered for RevOps, pipeline/marketing ops, and customer ops.

Revenue Operations

  • .
  • .
  • .
  • .
  • .

Pipeline & Marketing Ops

  • .
  • .
  • .
  • .

Customer Operations

  • .
  • .
  • .
Metrics Mentioned

The numbers, with context

$100k bet → $250k EV
Worked expected-value example

A 50% chance of returning $500k and a 50% chance of returning nothing. Used to show the arithmetic is simple enough that skipping it is a habit problem, not an analytical one.

~90% win rate; losing five in a row < 0.01%
Pocket aces versus seven-deuce

The boundary on process-versus-outcome. Below that threshold it is variance; at it, the correct conclusion is that the game is rigged rather than unlucky.

~10 minutes
Time to model a conference bet

Cost, travel and loaded time against target leads at a target ACV — enough for the decision, and explicitly not expected to be precise.

2–5x
Sales conversion lift reported by Tavus customers

Vendor-reported customer outcomes for the sales use case, cited by Yishi rather than independently measured.

60% higher engagement, 300% faster ramp, 54% higher confidence scores
Learning and development results reported by Tavus customers

Vendor-reported outcomes for the employee enablement use case, which Yishi describes as a common entry point before customer-facing deployment.

+23%
AI interview completion versus audio alone

Vendor-reported lift in completed interviews when conducted with conversational video rather than audio only.

Entities

Companies, people & tools mentioned

Auto-extracted and linked into the knowledge graph.

Companies

TavusAI Infrastructure

Where Yishi leads go-to-market strategy, roughly three months in at recording. Conversational video AI building PALs — Personified Application Layers — which he describes as the market leader in the category and as the back-end infrastructure behind many AI SDR and AI interview products. Customer-reported results span sales conversion, L&D engagement and ramp, and interview completion rates.

00:00 · 30:02 · 39:53Company →
Goldman SachsFinancial Services

Where Yishi began his career in the investment banking analyst programme in San Francisco, the first step on a finance track he describes as the default path for a hardworking student in the late 2000s.

03:04Company →
DeepBenchExpert Network

The expert network Yishi co-founded with MIT Sloan classmates in his first week of business school. It raised money, reached seven figures in revenue, became profitable at one point, and was acquired — the experience he credits with catching the entrepreneurship bug.

04:25Company →
LeanScaleGTM Operations

Anthony's firm. He uses it as the counterpart example on pattern recognition — an agency sees the same engagement problems repeatedly and goes deeper for it — and floats a discovery-agent use case where clients could complete research interviews with an AI at any hour instead of consuming the team's meeting time.

29:03 · 40:49Company →
OpenAIFrontier AI

Raised by Yishi when weighing competitive risk on his career bet: OpenAI attempted video with Sora and shut it down, and larger entrants may still come, which he treats as a known risk rather than a reason not to take the position.

35:24Company →

People

Yishi Zuoguest

Head of Go-to-Market StrategyProfile →
Anthony Enricohost

Co-founder of LeanScale and host of The LeanScale Podcast.

Co-FounderProfile →

Tools & software

Methodologies referenced Thinking in betsTen-minute ROI modelDecision post-mortem
Related LeanScale Services

Where this shows up in the work

AI for GTM

Agents and workflows that multiply operator throughput — the 'make humans superhuman' thesis, operationalized.

The episode's second half is a practitioner's view of deploying conversational AI in go-to-market, including why integration rather than model quality is the constraint.

Fractional RevOps

Embedded, field-facing RevOps leadership and execution — the operator model Jimmy describes, delivered as a service.

The ten-minute bet model and the opportunity-cost comparison are exactly the operating discipline an embedded RevOps team installs.

Forecasting & ARR Definition

Define a raise-worthy ARR and build a reliable consumption forecast across finance, data science, and go-to-market.

Expected value, stated assumptions and process-versus-outcome are the reasoning underneath a defensible forecast.

Frequently Asked Questions

Straight answers

Generated from the conversation, marked up for search and AI extraction.

What is expected value and how do you apply it to a go-to-market investment?

Expected value multiplies each possible outcome by its probability and sums the results. A $100,000 investment with a 50% chance of returning $500,000 and a 50% chance of returning nothing has an expected value of $250,000. Applying it to go-to-market means pricing a conference, campaign or agency the same way: what it costs including the time it consumes, against the outcomes you think it can produce and how likely each is. Two checks matter as much as the arithmetic — whether the same resource has a better use, and whether losing the money is survivable.

What is the difference between making a decision and making a bet?

A decision has broadly known outcomes: eggs or yoghurt for breakfast, water or coffee. A bet has meaningful variables outside your control that can derail the result — entering a new market, launching a product, funding a channel. Most business calls are bets, and the problem is that operators make them in meetings without pausing to identify them as bets or to price the expected value. Naming which one you are making is the step that changes the conversation.

How do you separate process from outcome without ignoring bad results?

Judge the decision by the reasoning available at the time, because a well-reasoned bet can still lose. The boundary is statistical: holding pocket aces against seven-deuce you should win about 90% of the time, so losing five all-ins in a row has a probability under 0.01%. At that point the conclusion is not bad luck but that something is wrong — the deck is rigged, or in business terms the process is broken. Defend good process against a single bad outcome, and investigate it when the losing run exceeds what variance can explain.

How long should it take to model the ROI of a conference?

About ten minutes. Add the direct cost — tickets and travel — to a rough estimate of the salary cost of the people attending, then set against it the number of strong leads you expect and the ACV you are targeting. Three or four significant figures is close enough. The output may be wildly inaccurate, and that is acceptable, because its purpose is to let you compare this use of a week against the outbound calls or campaign the same week could fund.

Should marketing spend be modelled when attribution is unreliable?

Yes. Attribution is fundamentally harder on the marketing side than the sales side, but that makes an assumption necessary rather than optional. A $100,000–150,000 rebrand can be reasoned about in thirty seconds: what probability does this add to raising the next round, and what is that worth? The assumption may be wrong and can be revised as evidence arrives. An absent number cannot be revised, and it removes the spend from comparison altogether.

Should a beginner work on reading people or on fundamentals?

Fundamentals, and by a wide margin. Mastering the basic disciplines makes you better than roughly ninety-five percent of beginners, and reading tells is a genuinely advanced skill that does not matter until the basics are automatic. The poker example is how you hold your cards when you look at them — trivial-sounding, commonly done badly, and a question of discipline. The business equivalent is that time spent on subtle buyer signals while the basics are unreliable is taken from the work that would actually improve results.

Is choosing not to act a neutral option?

No. Making no decision is still a decision and carries consequences: declining to enter a new market means a competitor may take it. Inaction tends to escape the scrutiny applied to action because it looks like caution, so its cost should be put on the table alongside the alternatives and made to win on merit.

What is holding back adoption of conversational video AI inside companies?

Integration capacity rather than the technology itself. Tavus's customers are currently weighted toward AI-native companies that can build on the API, because an AI human only becomes useful once it is connected to a knowledge base and to surrounding systems such as HR — otherwise it does not know what it does not know. That scaffolding requires AI engineers most companies do not have in house, which is why learning and development, deployed internally first, is a common entry use case.

Timeline

Jump through the conversation

Click any chapter to open it in the transcript.

00:00Cold open + intro 02:24From Goldman to a hedge fund to founding DeepBench 06:52What finance actually teaches you about GTM 08:21How poker got its hooks in 11:03Winners vs. losers: goals, discipline, fundamentals 14:02Expected value, explained 17:03Process vs. outcome (and pocket aces five times in a row) 20:01A decision vs. a bet — the distinction most operators miss 24:00The Hail Mary: when the right call still loses 27:35Good player vs. great player: reading the tells 30:30Thinking in bets at Tavus: conferences and outbound 33:22The career bet: why he joined Tavus 36:30Where AI humans go next: digital twins, senior care, enablement 39:53The ROI numbers customers are seeing 40:49A discovery-agent idea for LeanScale + wrap
Full Transcript

The whole conversation

Broken into chapters, searchable, verbatim from the audio. Speakers inferred (not diarized).

Transcript as markdown ↓ Whole episode as markdown ↓
00:00Cold open + intro

0:00 Our guest today is Yishizuo, who leads go-to-market strategy for Tavis, the conversational video AI company, Building Pals, which stands for personified application layer.

0:12 These are emotionally intelligent AI humans or interactive avatars.

0:16 They actually like didn't have an open role, but they had some needs and, you know, through the interview process, I was like, hey, I can do this.

0:23 They saw what I could bring to the table and they kind of created this role for me, like head of go-to-market strategy, you know,

0:29 do some simple research. It's very clear Tavis is at the cutting edge.

0:33 They're actually the market leader in conversational video, so like AI avatars, they can talk back to you.

0:39 There's a lot of feel good use cases that I think will pop up and I'm really interested to see how it unfolds.

0:45 The football analogy, there's that famous play like the Seahawks versus the Patriots, like 10 years ago, where they went for--

0:52 Yeah, hand the ball to Marshawn Lynch.

0:53 Yeah, yeah, yeah.

0:54 Well, you know, look at the-- I mean, I love football. I don't--

0:59 Yeah, yeah, so I mean, it's--

1:01 Along with your professional background, you've also spent a lot of time in the poker world.

1:08 You have to separate the process from the outcome.

1:10 Let's say you think you're following the right process and you get bad outcomes every single time, and maybe you need to reevaluate your process.

1:17 I'm going to use a poker analogy.

1:18 Let's say you have pocket aces and the other guy has seven deuce, best handed in the world versus worst handed in the world.

1:23 You keep getting aces and the other person keeps getting seven too and you keep going all in.

1:27 You should win like 90% of the time, but somehow you lose like four or five times in the world.

1:32 Let me ask you, like what do you think is happening there?

1:34 In a certain industry, what are those three points?

1:37 Yeah, I'd say, you know, we talked about like thinking bets, like expected value.

1:42 I think that's probably the most applicable like framework in from poker to a lot of like business logic.

1:48 You can kind of see like this is where the future will inevitably head and I'm making a bet.

2:00 You've been a banker, a hedge fund analyst, a founder, head of finance, and now you're running go to market.

2:06 How did you go down such an eclectic path?

2:10 And what's the through line?

2:12 Thanks, Anthony.

2:13 I guess I'll start from the beginning.

2:14 So I'm an immigrant to this country.

2:16 I was born in China.

2:17 I moved here when I was six years old and I grew up in the San Francisco Bay Area with my parents.

2:22 I'm an only child.

2:23 It wasn't always easy.

02:24From Goldman to a hedge fund to founding DeepBench

2:24 Like my mom was a waitress for many years when we were growing up.

2:27 And, you know, my dad was a software engineer and like started in like 1999 right before the dot com bubble.

2:33 He was laid off a couple of times in the early 2000s.

2:36 So like growing up, like it was it was challenging.

2:38 Like money was always a challenge and going into college.

2:41 I was like, I'm going to know.

2:42 I don't know what I'm going to do, but I'm going to try to make money.

2:45 So, you know, in their late 2000s, like 20, 2007 to 2011, when I was in college, it's like this is before the big tech companies really got really big.

2:55 And the cool thing to do, I don't know about cool, but it was like the thing to do if you're like, you know, reasonably intelligent, hardworking kid was try to go to investment banking.

3:03 And that's exactly what I did.

3:04 I went to Goldman Sachs in San Francisco, IBD analyst program, pretty typical.

3:09 I'm sure many of the listeners of know people have gone through that program themselves, had a great experience, but I really no idea what I want to do.

3:17 Like I kind of made it into corporate America from this like poor, like hungry immigrant kid.

3:21 I'm like, what's next?

3:22 And I did.

3:23 You know, the next natural step was to go to a hedge fund because that was a natural path along the finance track.

3:29 I went to a small hedge fund.

3:30 A lot of great people spent three years there.

3:32 It was a good learning experience, but it just wasn't the right fit for me.

3:36 I'm a very, I'm an extrovert at the, when you're an investor reading 10 K's, it would be like days before I would talk to anyone besides my coworkers.

3:44 And just in the long run, it wasn't the right fit.

3:46 And I kind of knew it at the same time.

3:48 This is like 2013, 14, 15.

3:50 I was in San Francisco.

3:52 All these like startups were like popping up and I'm like, you know, these people are building these startups.

3:57 Like, you know, what do I have to lose?

3:58 Like I let me give it a shot.

4:00 And I actually try to build some startups in my nights and evenings back in the day.

4:04 I just fell flat on my face.

4:06 I had no experience.

4:07 And I was like spending money on freelancers, just doing every single thing wrong in the book.

4:12 But I kind of like caught the, caught the bug.

4:14 I, you know, want to do more of that for whatever reason.

4:17 So I went off.

4:18 I was going to business school at MIT Sloan.

4:20 Like the first week of school, found some classmates and we started iterating on ideas.

4:25 We ended up building Deep Bench, an expert network.

4:29 There's articles written about us in Forbes.

4:31 Four MIT students built this thing and we did build this thing.

4:34 Raised some money along the way.

4:36 Got to seven figures in revenue.

4:38 Became profitable at one point and had an exit.

4:40 And that was a thrilling ride.

4:42 That was a lot of fun.

4:43 I stepped away from that in like late 2020 and I wanted to like, you know, figure out what's next.

4:48 And I still didn't really have any idea.

4:50 So I had done, I was like deeply ensconced in the world of entrepreneurship.

4:55 And I decided to just, you know, keep building companies.

4:57 I started building a startup studio with, ended up joining forces with a partner.

5:02 And we built a few businesses.

5:04 It was a good experience.

5:05 And I did that for about four years.

5:07 All sorts of different like software enabled businesses.

5:10 Had some profits here and there.

5:11 But ultimately kind of like towards the end of 2024, I realized like this isn't really like what I want to do.

5:17 I want to do something bigger.

5:18 So I decided to go back into, you know, working for someone else for the first time in many years.

5:24 Ended up joining like a YC back healthcare AI startup.

5:28 It was a good ride.

5:29 A challenging thing is like the role we kind of figured out was finance because like, you know, Goldman Sachs hedge fund.

5:35 And I did that plus a bunch of other things because there's a series A startup that you have to be a generalist at that stage.

5:40 And it just wasn't the right fit for me.

5:42 I wanted to be more customer facing, like really like, you know, bringing in deals and all that kind of stuff.

5:47 That's what gets me going.

5:48 And I ended up stepping away from my previous company earlier this year and around February, March timeframe.

5:55 And I was introduced to Tavis by a friend of mine who was an early investor that they actually like didn't have an open role, but they'd had some needs.

6:03 And, you know, through the interview process, I was like, hey, I can do this.

6:06 They saw what I could bring to the table and they kind of created this role for me, like head of go to market strategy, which in short, I like help figure out who to target, how to target them.

6:16 And I do the, like, you know, the execution work beyond strategy, too.

6:20 I'm like going to conferences, I'm hopping on sales calls.

6:22 This week I've been working on a brand redesign, like managing agencies, all sorts of stuff.

6:27 So, yeah, that's kind of me in a nutshell, kind of how I got here.

6:33 You know, there's been a handful of people that I've worked with or interviewed that have gone through a finance track and then found their way to go to market.

6:42 Is there anything that you think you learned that is helping you be more effective in a go to market role by going through those roles and being in that industry for a little bit?

06:52What finance actually teaches you about GTM

6:53 Yeah, I think inherently, I don't know if it's the fact that work in finance and I may just be, I'm sure it actually had something to do with it, but I am like relatively analytical.

7:02 Like we were constantly running, any company runs experiments and go to market, whether it's like this conference, this campaign that I tend to be, at least I find myself at my company, a little bit more kind of like ROI focus.

7:14 Like, hey, like if we're going to invest this much resources, this much time, like what is the expected ROI?

7:20 Just like that muscle, that way of thinking and that rigor.

7:24 I think I bring that to the table and chalk it up to my finance background or just, you know, who I am.

7:29 But I like to like, I try to be rational about like how I place my bets.

7:34 Yeah.

7:35 And a lot of people tend to see the game of go to market as a little bit more subjective, a little bit more art than science.

7:43 But the more time I've spent in go to market, the more it feels like a science there, plays to be made, motions to be run, data to be analyzed and then bets to be made on the data that you have access to.

7:58 Definitely, definitely.

8:01 One thing when we were prepping for this, I found it fascinating.

8:05 And I'm really, really excited to go into it is I'm not sure where you're fitting this in.

8:10 But along with your professional background, you've also spent a lot of time in the poker world.

8:17 And I think there are so many parallels.

8:20 But how did you get into poker?

08:21How poker got its hooks in

8:22 And what is what is the allure that draws you to that game?

8:27 Yeah, good question.

8:28 So I think I first came across poker, you know, like a lot of people probably listened to this back in like 2003, four or five, like, like Chris Moneymaker, like on ESPN won the World Series of Poker.

8:39 And there's this big poker boom.

8:40 I was a little bit too young to like really start playing poker back then, like seriously.

8:45 But it got me and my high school friends start.

8:47 We start started to play like, you know, Friday nights, pizza and video games.

8:50 That's what we did.

8:51 And we were the cool kids.

8:52 Not really.

8:53 But it was fun.

8:54 And I honestly, playing with my buddies back then, I was not one of the best players.

8:58 Like, I was I want a couple, you know, everyone wins here and there.

9:01 We were playing for like 10 bucks, by the way.

9:03 Everyone puts in 10 bucks and like someone's mom would buy us pizza.

9:06 It was that's what it's like.

9:08 And I didn't really play for like, like seriously.

9:11 I play like once a year, like on average over the next like 10 to 12 years.

9:15 And then in 2019, towards like, you know, the middle while I was working my startup, I was like in Boston, like a lot of my friends had moved away after business school.

9:23 Like they opened up on on core Boston Harbor and then I just went like once or twice and I kept winning.

9:29 I'm like, wow, like poker is easy.

9:31 And then, you know, that kind of got me like really into it.

9:34 I learned a few months thereafter that I was just not as good as I thought.

9:39 But, you know, I kept at it.

9:41 It became like a hobby that, you know, when it's a hobby that helps you make money and it's like fun and you're good at it.

9:46 And like it just becomes a thing.

9:48 And, you know, I do other things in life too.

9:50 But like it's nice to have something that you can really like put your efforts in and be good at outside of the workplace.

9:55 So kind of like it's like a nice like competitive release for me.

9:58 It's fun.

9:59 I get to meet a lot of people.

10:00 That's kind of how it started.

10:02 And yeah, like I over time, like after I took a step back from deep bench, I had a lot of freedom in my life, a lot of optionality.

10:12 Like I was like exploring businesses, building things.

10:14 I had full control over my schedule, almost like too much like freedom in some senses.

10:20 That's another another another story.

10:22 But I, you know, I also like, you know, in startup world, like money comes and goes, right?

10:27 It was nice to have like a hobby that can like, you know, help pay the pay the bills from time to time.

10:31 It was that's kind of how it evolved.

10:33 Like I love the game.

10:34 I don't play it like as seriously as I did a few years ago, but I still like love it.

10:40 Well, coming from someone who has probably played three casual games with fake chips in his life.

10:47 So let's just say I've never really played.

10:49 What is the separation between someone who is consistently winning in poker versus people who are constantly losing?

10:59 Where's where's the skill build as you become more professional at the game?

11:03Winners vs. losers: goals, discipline, fundamentals

11:05 Yeah, good question.

11:06 I think I think we have to start kind of at the very beginning kind of like what is you have to understand like what is your goal?

11:14 For example, like for some people like you don't play poker necessarily to make money.

11:19 It's to have fun or it's like you're playing at a poker game with a bunch of prospects who could become potential clients.

11:25 I'd argue it's worth losing a few hundred bucks to, you know, build relationships that could lead to a few hundred thousand and of ACB ARR or whatever it is.

11:33 Right. So I think that that's a key part of it.

11:35 The second part of, you know, poker, I think that's applicable, like across different disciplines.

11:40 It's actually discipline.

11:41 And like that's how I describe it.

11:43 There are certain like basic things you do like in poker that, you know, may not matter that much in the beginning.

11:49 But like if you don't work on it, it's like the fundamentals of like learning how to dribble basketball or like or like practicing your golf swing.

11:57 There's certain fundamentals in your in your game and to work on that.

12:00 One basic thing is like how you look at your cards when you pick them up.

12:03 You want to hold them in a certain way such that like no one sees it.

12:06 And it sounds so basic.

12:07 But, you know, beginners just aren't good at this.

12:09 And it actually takes discipline to learn like how to do it the right way.

12:13 And you might learn the hard way how how not to like look at your cards.

12:17 But it's like poker ultimately and in life, it's a lot of these like little things that you want to become a master of.

12:22 You just have to get better at it.

12:24 And then once you start mastering kind of the fundamentals, other elements like, you know, some of the more advanced stuff like reading people, giving off tells.

12:32 It's actually very advanced.

12:33 Like if you work on the first layers of fundamentals, like you'll be better than like ninety five percent of beginners.

12:39 Like don't even worry about like tells and reading people.

12:42 It actually doesn't matter until you get more advanced.

12:44 So I'd say work on figure out your goals and like work on the fundamentals by being disciplined, which matters for a lot of other disciplines outside of poker, too.

12:53 Yeah, I think being exceptionally well at the basics is how to be exceptional.

12:59 And there's so many times where people focus on the fancy aspect rather than like, hey, let me just do an incredible job at the basics.

13:10 And if you do that, like you said, it usually separates you from everyone else.

13:16 In this case, how how do you relate some of the things happening in a poker game to go to market?

13:26 What are some of the parallels that maybe shape the way you think about how you approach building a business, going to market in a certain industry?

13:35 What are those three points?

13:37 Yeah, I'd say, you know, we talked about like thinking about expected value.

13:41 I think that's probably the most applicable framework in from poker to a lot of like business logic.

13:48 Because ultimately, for those just for those who might not know exactly what that means or exactly what that is, how would you explain expected value to someone new?

13:57 Oh, yeah. So let's say you you're in business, there's like two projects.

14:02Expected value, explained

14:02 One is like you invest a hundred thousand dollars, but it has a 50 percent chance of returning five hundred thousand dollars and a 50 percent chance of increase returning zero.

14:12 So the expected value of zero times zero or 50 percent times zero plus 50 percent times five hundred thousand is zero plus two hundred fifty thousand.

14:21 So you're a hundred thousand dollar bet has a, you know, expected value of two hundred fifty thousand.

14:27 Now, like, should you take that bet is the question.

14:30 Well, there's a lot of factors that go into that.

14:33 One is like, what are the other options?

14:35 Maybe you're a hundred thousand dollar bet. There's like four other opportunities that have expected value of eight hundred thousand or something like that.

14:43 What if it's the last hundred thousand that you have?

14:45 Like you're going to go homeless if you spend this hundred thousand.

14:48 Well, like you can't take a risk where you have a 50 percent chance of going homeless.

14:51 You might take a bet that has a ninety nine percent chance of returning one hundred twenty thousand versus like, you know, like it's just a matter.

14:59 There's a lot of factors that go into play.

15:00 There's like the opportunity cost of like what you spend your money on.

15:04 And I mean, this this this framework is kind of very applicable to poker and poker like you don't really know.

15:11 You're not sure you know what your hand is, what the probability is of like getting certain things.

15:17 You're not quite sure what your opponent has.

15:18 There's a lot of layers of like game theory to this, but fundamentally it boils down to expected value and whatnot.

15:28 So you see you have decent hand. Maybe it's likely they have worse hands.

15:35 Then you know, like, OK, because the likelihood, maybe this is a time where it's good to double down.

15:41 Or if you feel like it's more likely they have a good hand, less likely that you have a good one, then you can start to kind of hedge your bets.

15:48 I know there's also like different bluffing strategies and things like that in poker.

15:52 But just in the expected value realm, it's a good way to evaluate what you should do.

15:58 Like, should you double down on something? Should you pull back?

16:01 And then also knowing, hey, it was the right bet to make like the likelihood of you getting an outsized return was higher than not.

16:11 So it was the right bet to make. But sometimes they don't pan out and that's OK.

16:16 And I think that's something where people have a really, really tough time doing like being at peace with making the right bet, but the bet not panning out.

16:28 And do you build that muscle by playing poker, you think?

16:34 Oh, yeah, I would say so. It's it's it's very this is an area we could go like pretty deep in like there's this idea of like process versus outcome.

16:43 This is like fundamental to like probability, like investing, like trading, definitely poker and just business decisions, in my opinion.

16:53 You have to separate the process from the outcome.

16:55 But also, like, let's say you think you're following the right process and you get bad outcomes every single time and maybe to reevaluate your process.

17:03Process vs. outcome (and pocket aces five times in a row)

17:03 Maybe there's some like a variable that's coming out, let's say, like, I'm going to use a poker analogy.

17:08 Let's say you have pocket aces and the other guy has seven dudes, best hand in the world versus worst hand in the world.

17:12 You keep getting aces and other person keeps getting seven to and you keep going all in.

17:17 You should win like 90 percent of the time.

17:19 But somehow you lose like four or five times in a row.

17:21 Let me ask you, like, what do you think is happening there?

17:23 Like this just as a late as someone who doesn't know what do you think is happening there five times in a row?

17:29 I mean, to me, I would just take that as bad luck.

17:32 Like if I know that this is a really strong hand and the likelihood of somebody having a stronger hand is really low.

17:39 And in this case, the game is known.

17:42 Then I would say keep doing it.

17:45 But that's as a beginner, maybe not knowing if there's another strategy to take.

17:49 Well, so this is where like you have to like reexamine process as an outcome.

17:53 And if you lose five times in a row, I think the odds aces versus seven dudes off.

17:58 It's like somewhere like it's between like 10, like nine to one or something like that.

18:02 It's eight to one or seven to one or nine to one, something like that. The probability losing five times is somewhere like less than zero point zero one percent.

18:10 It's somewhere very low. So you're probably in cheated. That's what it is. Right.

18:14 So you have to reevaluate kind of like what's happening.

18:16 Like the deck is rigged. Something's going on.

18:18 Well, I wasn't. Yeah, I wasn't maybe allowing that case.

18:21 But yeah, I think something you would think something is off.

18:25 Yes. So so this is this is kind of going back to for the listeners applying it to like the business world, like thinking about like like you have to.

18:34 I believe that we should have like a mental model of like if we're going to invest as much money in this area, taking a bet, like let's actually think through the assumptions.

18:43 And a lot of times you just don't know you have to make reasonable assumptions.

18:46 But that's not that's not an excuse not to come up with assumptions.

18:49 And sometimes like especially on more on the marketing side versus the sales side, it's attribution is like just fundamentally hard.

18:55 We should still have like a model of like, you know, what is the expected spend of this like brand redesign?

19:00 Like we should have that number somewhere. Right.

19:02 It's like it doesn't you don't have to spend like five hours on you could spend like five minutes or like 30 seconds.

19:08 I'm like, here's like we're spending one hundred thousand hundred fifty thousand on like a website read as rebrand.

19:13 Like what is the expected value? OK, it's going to increase the chance we're going to raise our next series see a hundred million dollar round by ten percent, twenty percent.

19:21 OK, that already that's all the math you need for like some of this stuff. Right.

19:24 But you're still still trying to think in terms of numbers, if that makes sense.

19:28 It does make sense. And I think something that's really important for an operator, a founder, somebody who's running a go to market team,

19:37 really anyone in in business in general. But can you make the distinction between making a decision versus making a bet?

19:50 Yeah, I mean, I mean, you decide to make the bet that that's that is part of it.

19:55 And I think when you make a decision, there are certain outcomes that you believe are associated.

20:01A decision vs. a bet — the distinction most operators miss

20:03 Like you're implicitly saying, if I do this, that will happen. Sometimes like nothing in life is guaranteed.

20:09 Right. Like a meteor could hit us tomorrow and like you never know. Right.

20:12 But the idea generally with most decisions, like you make a decision, you have an idea of what will come.

20:17 And that probability might be ninety nine point nine percent. And in fact, like that's actually a great question.

20:21 I actually haven't thought about that. I think most decisions are implicit bets of some sort, if not explicit.

20:28 Yeah. And I think one one way to frame it based on what you said is if you're making a decision, the outcomes are relatively known.

20:37 I'm making a decision to have eggs or yogurt for breakfast. I'm making a decision to drink water or coffee.

20:45 And there's no real probability of something happening. You're making a decision.

20:51 But when you're building a business or running a plan, you're making a bet. There's things that are completely outside of your control that can completely derail your plan and things that can happen that are unexpected.

21:03 So you have to know when you're going down some when you've decided to make that bet, understanding the risk and evaluating whether it's the process or outcome issue.

21:16 So I think a lot of times people are in meetings, having calls, making decisions on things and not realizing that they're actually making a bet and taking the time to evaluate the bet that they're making and taking the time to process what the expected value of that bet is.

21:35 I don't think a lot of business people and operators are really taking it to that level before making a decision on what direction to go.

21:44 Yeah, I'd say and I think that's okay for reasons I'll explain, but let me take a step back.

21:49 So what you talked about, like, you know, waking up or like getting eggs, like whatever, drinking milk, it becomes like habitual almost.

21:55 You don't think about it. And that's that's what I was saying earlier, like some of the good habits in poker, like learning how to look at your cards the right way, you don't make a decision after certain becomes like muscle memory.

22:04 And it's like just, quote unquote, the right thing to do, right? Like it's it's like, you know, learning tennis or basketball, like doing things the right way. It becomes so it becomes so ingrained in your muscle that like it's not a decision anymore.

22:16 And you basically remove that uncertainty, if you will, it becomes part of your process. So some of these, like, you know, you don't you can you don't need to apply the same analytical framework, like to every decision.

22:28 Like we just don't have, you know, there's this thing, like mental burden, like mental, like you get tired, like, if you always think like this, but if you if you do a little bit of it, like, it just becomes natural, like, and you sometimes you just really don't know, like, there's some sometimes really tough questions, like, like, you know, these strategic decisions, like, how do we really like know for sure?

22:46 And you just you just don't like we're talking about entering new markets like this week, like, how do we know that there's so many uncertainty, like, you can make all the probability models in the world, but someone ultimately just needs to like, okay, like, we know enough.

22:57 At a certain point, you just have to commit. So that's, that's what it is. And like the beauty of poker, and you know, we're talking a lot about that is it kind of like it forces you like there's all these decision points.

23:06 And you just have to be you have to embrace the uncertainty, like you have to be okay, like, losing sometimes, even when you have the upper hand, and you just realize, like, look, look at your process, unless you're losing like five times in a row, like, the game will, you know, if you're playing it the right way, the game, the result will take care of itself if you do things the right way.

23:24 Yeah, and I think the issue I tend to have is when people call people out when something doesn't work. So, hey, we went into this new market, oh, it didn't work. So, and then they take that a step further and be like, Oh, well, you're a bad operator, or you don't know what you're doing.

23:41 Or it's like, Hey, whoa, we nobody knew the outcome of this, we laid out all the information we had at the time, it was the best decision to make, you know, I think it's, hey, it's the end of a football game, and you're down by a touchdown, and you have to throw a Hail Mary, like, that's the only play to make you have to do it.

24:00The Hail Mary: when the right call still loses

24:01 It's not like there are other options. And then it's like, Oh, well, that was a bad play, because it didn't work out. And I think people just don't really think, I think my experience, people don't think deeply enough about all the variables when they're making big decisions like going into no market launching a new product doing XYZ.

24:21 There's also making no decision is a decision too. So doing nothing is a route to take, and that has consequences as well. If you don't go into that new market as a competitor, go and take it.

24:32 So either way, like you have to make a call and do something you have to make a play, you have to double down fold, whatever. And I think in the long run, you should start to rack up the wins, but you have to be okay making it through those losses sometimes.

24:49 Yeah, agreed. I mean, the football analogy, there's that famous play like the Seahawks versus the Patriots like 10 years ago, where they hand the ball to Marshawn Lynch.

24:59 What are you doing? Yeah, yeah. Well, you know, look at that. I mean, I love football. But the whole defense knew that that's what's going.

25:05 Yeah, yeah. So so I mean, it's it's arguable. But, you know, back to the point of like, you know, when something goes wrong, like, you know, like, how do we how do we think about how do we do like retros like postmortems, all that kind of and this is the stuff like good companies think about.

25:18 Like, when I was at a hedge fund, like, like, really, like, it was all about decision making. We read books and books, like, about decision making. That's all it is, like, making good decisions, avoiding cognitive biases.

25:29 I think that actually has trained me well. And like poker has also trained me well to like think this way. I think, you know, like, back to like, you know, how do we do postmortem, like, it just gets in real life, it gets complicated because there are these like other factors.

25:43 Let's I know, I think you guys work with clients, right? So like, there's there's like, you know, like, you know, sometimes you're not able to communicate that you have to be delicate of like, how you communicate it, even if you are the client within the company, there's like internal politics.

25:55 But you have to like, you just have to balance that with like common sense. And, you know, it's every situation is a little bit different, like, knowing like what to do and actually doing it.

26:03 There's like so many things like what neither knowing what to do isn't always easy, even if you know what to do, like doing it, there's all these like, you know, humans like, you know, we need to solve for that factor, right?

26:13 So it's not always easy.

26:17 It's not easy at all. And I think the people who are in the game long enough, like you said, you build that muscle memory, you start to see the patterns.

26:26 And then if you stay disciplined in your decision making, tuning down your cognitive biases and running the process that should give you the best outcome in the long run, you should see the results.

26:39 One thing, mainly just because I'm curious. So getting the fundamentals down makes a lot of sense. What are some of the things that separate a good poker player from a great professional poker player?

26:52 What are some of the things that the best in the world can do that amateurs find it very difficult to nail down?

27:00 Yeah, good points. And you kind of alluded to this like when you're talking about like pattern recognition and whatnot.

27:06 Like once you mastered like the basics and you've gotten the reps in like, you know, whatever Malcolm Gladwell's 10,000 hours, like whatever it is, you play 10,000 hours of poker.

27:16 You've seen a lot of like just basic situations and you become very attuned when you notice things that are out of the blue.

27:23 Like there's all these like signals that are constantly like being like sent to you, like you just pick up on it.

27:30 I think Dan and Agranu has like some like thing where he talks about this and like I'm not the biggest fan of him, but like he has some good points.

27:35Good player vs. great player: reading the tells

27:37 If you're like a good poker player, like the better you are, the more things you notice that like deviate from the norm and very quickly like hone in on that.

27:45 What is what is different about this? Like why is that different? Could it be could it be something innocuous?

27:50 Let's say this guy is like all of a sudden very alert. Like could it be that he has a good hand or could it be he just like chugged a cup of coffee or like, you know, like his like wife just came over and like said something like very dramatic.

28:03 I don't know some something it could be like what is actually driving that and you kind of dig and you start paying attention more. The thing is like there's just so many signals. If you're a beginner trying to work on that, like good luck to you because like if you don't have the fundamentals of like.

28:16 Like like position and like holding your hand properly. If you don't have that down, don't even think about like thinking about like picking up things like minuscule tells you're going to you're distracting yourself from like working on the things that are going to be really helpful.

28:29 And you know like connecting it back to like the business world. It's like like you you jump on like a thousand sales calls ten thousand sales calls over the lifetime of your career.

28:39 Like you just get better at it. If that's what you're doing right. If you're like reviewing like marketing copy for like landing pages. This is why agencies exist like they just have seen so much of it. Obviously like they don't have as much domain like company expertise on your market.

28:53 But it's it's like that muscle memory like it just like in poker just like in like professional services. It really really comes into play.

29:03 Yeah and as a agency owner, it's definitely the nuances that come up in every single engagement. We see the same things over and over and over again and you can just go a little bit deeper where if you're doing it for the first time like there's no way you're going to pick up on certain things.

29:22 Because there's too too much and especially when you know like hey I have put in the ten thousand hours and I'm still picking up new things and I'm still adding things into my arsenal of things to look at and then recognize during during any engagement.

29:40 I think that's when you also appreciate like how far you can go within any discipline to for sure.

29:48 What's something at Tavis that you've applied this thinking in bets mindset to a real go to market call you've refrained as a bet and aim the company towards it.

30:02 Yeah so in terms of things that you know I'm like only three months and by the way so I'm like applying this like wherever I can one a couple areas where I'm very focused on is like attending more conferences and two we're starting to work with external like outbound lead generation agencies.

30:19 And very much so like we structured in a way or at least either structured or think of it as like that it's like every single investment I'm like this is like the outcome that will probably come from it and there's there's a lot of assumptions that need to be made like if we go to a conference for example this is like how much we're spending we're bringing these people.

30:30Thinking in bets at Tavus: conferences and outbound

30:37 The investment of like how much we're spending for the conference tickets the travel cost that stuff's easy it's just like a number we don't need to get super exact it's like within like you know three significant four significant figures within like about plus or minus a couple thousand dollars that's close enough for our purposes.

30:53 And then like you know plus our time as well which you know rough estimate of salaries and you just really need to account for that and then go to this conference and what's the expected outcome like we think we can get at least like you know 10 like strong leads or like

31:08 20 whatever it is like what is the ACB that we're targeting then you just build like a very simple model it literally takes probably 10 minutes and it may not be accurate could be wildly off hopefully in the good direction but at least like we tried right like at least like we're thinking about it this way.

31:22 So I think like other than me and not a lot of people like I think our founder probably implicitly thinks this way I'm trying to make some of this a little bit more explicit.

31:31 Well I think also when you do that then you have to evaluate the opportunity cost like your time okay I'm going to be at this conference for a week but if I was here I could have made X amount of cold calls or worked on this other campaign and if you start to apply that mindset even if even if it's not accurate or very precise it can still help you make the decision between two different things you can apply your time and resources towards.

31:59 And maybe the magnitude could be off but at least it can help you choose between two things and you have a better idea that this will create more value than the other one and if you keep stacking those decisions up over and over and over again and then in later stages you do have more data to make better bets then that compounding effect can really drive success.

32:25 Yeah that's right it's just a muscle like you'll get better at making these calculations questioning assumptions and your assumptions will just get better too like it's like a nice forcing function to like really think about like what are we doing here and why we're doing it.

32:41 So you've also made a pretty big bet on your career joining Tavis powering AI interviewers SDRs everybody's talking about it. What did you see in the market that made you make this bet on your next step in your career.

32:58 Yeah I mentioned my friend investor with seed stage investor this company is now series B he's done well and he just told me about this opportunity and I was like looking for my next thing and had a conversation I was very convinced by the founder's vision of like where this is heading

33:15 and you know do some simple research it's very clear Tavis is at the cutting edge they're actually the market leader in conversational video. So like AI avatars that can talk back to you.

33:22The career bet: why he joined Tavus

33:26 I did a little bit of playing around the product on my homepage ishyzo.com go to about you can actually talk to a video avatar of me it'll like talk back to you.

33:34 I use like a crappy MacBook Pro camera it's like the quality could be better it would be better if I went to a professional studio but like it's pretty good you get a sense it's a demo but it's the technology keeps getting better and if you think about the arc of like human history

33:48 this is part of like our companies like spiel like you know I've kind of I believe in it now like look at the arc of like how computers have evolved like you look at like you know science fiction it's always like you know Star Trek it's like you're talking with like a computer it's all like voice like you're

34:04 interacting with like 3D like holographic like avatars we're certainly not there yet but like you know we have to the technology has to meet where humans humans have to meet where the technology is and vice versa so we start out with like these like you know

34:18 graphical user interfaces like back in the day and now there's like multimodal and like we're starting to enter like AI has evolved to the point where you can actually like talk with the computer like a humanized version of a computer.

34:31 So what we do is like we just make it feel very natural we spent a ton of money like you know on AI research to get it to you know where it's still hasn't passed Turing test like you can tell it's not human but it's like far ahead of anyone else so like we're the leaders and like this field like all the basically

34:47 all the AI SDR companies some of your clients probably use some of them they are all they are all powered by our technology on the back end. We do a lot of like AI interview like we're the back end infrastructure.

34:58 Like like like customer training like customer support customer support training human resource onboarding we're doing stuff in like senior care like companionship as you imagine AI interactive humans are very helpful for that.

35:11 So there's just like this big opportunity if you believe in science fiction like I do you can kind of see like this is where the future will inevitably head and I'm making a bet like there are we do have smaller competitors like who knows what the future might bring.

35:24 You know open AI has tried video with their Sora shut that down you know they might come but this seems like a pretty good risk adjusted bet for me and you know like people are good I'm having fun I'm doing what I love like I'm like constantly thinking of experiments where I can like

35:38 you know target more customers and like you know running it from end to end have a lot of like a lot of freedom but also like I go in the office I have a really good culture here so it's a lot of fun.

35:49 What do you think are going to be the biggest applications you mentioned a few AISDR those are those are in flight right now, but if you were to maybe to rank your top three like a three five years from now this is just going to be ubiquitous people will interact with AI video avatars for these things.

36:08 Yeah, I think that digital twin space is going to be interesting like a few years ago like these like virtual influencers were very new like this this is all the rage and now there's, it's kind of like the hype has died down but like there are dozens I did a research quick

36:25 research the other day of these like virtual influencers out there I think a lot of fans to our knowledge like a lot of these companies like their fans want to talk to these virtual influencers but the technology like we're in the lead like we are starting to like work with some

36:30Where AI humans go next: digital twins, senior care, enablement

36:39 of these organizations I can't talk too much about that but they're like, there's gonna be virtual twin digital twins of like a lot of celebrities like you know CEOs creating twins of themselves to talk to their employees and like their employees can talk to them that that's what we do that's the key part.

36:54 That's going to be like an area that I'm very excited about. I think like companionship and like senior care like it's not enough caretakers like senior older people are lonely and it's a great use case for us it's a feel good use case to.

37:09 And then, you know, our core is like the B2B use cases like training sales people training people learning and development is like just overall across all businesses like a lot of companies like they want to put AI avatars in front of customers but maybe they're like,

37:24 they're like, Oh, we're not quite sure let's start with our own internal people first so we like learning and development as like an entry use case and we do have a bunch of use cases where these AI avatars already like interacting with with the end customers.

37:40 I love it. I think the digital twin one is really interesting, especially for family. You know, if you have maybe family members that have passed. If you kind of had a repository of who they are how they would have answered questions someone interact with if you want to bring something to someone.

37:58 I don't know, I think there's a, I think you're right. There's a lot of feel good use cases that I think will pop up and I'm really interested to see how it unfolds.

38:07 Yeah, I think we're a company is like right now like a lot of our customers are like an AI native companies that build on top of the API is what we're seeing right now in the market is that there's a whole universe of customers that may not have the engineers in house ready to like

38:22 get their APIs and build out like all the scaffolding that's required for like AI human to be useful because you have to imagine. Let's say you wanted to build an AI avatar that your employees can ask questions about like how your how your company like works or potential

38:38 customers can ask questions to make it useful you got to connect it to the knowledge base right otherwise the AI doesn't know what it doesn't know. So that scaffolding connecting like this to your knowledge base plus like you know connecting it to like other systems HR systems like

38:52 whatever it is requires like AI engineers like we are now starting to like build out some of that muscle so that we can work directly with end customer to deliver what they want so like a shout out to if anybody is interested like we can help you do that so come check out our technology.

39:09 I love it well a lot of the audience is interested in AI most of the people listening will be in B2B SaaS or AI themselves so and definitely on the forefront of leveraging whatever the latest AI use cases are for sure.

39:26 So a lot of these things for these companies tend to be I'll call them AI experiments very very cool very interesting and intellectually impressive that you can do these things but at the end of the day a lot of these have to have a real business impact.

39:44 Where are you seeing your technology driving ROI and driving business impact for the developers and people leveraging it.

39:53The ROI numbers customers are seeing

39:53 Yeah great question Anthony. We're seeing ROI across every single use case essentially that's quite incredible. So for the sales use case some of our customers customers are seeing 2 to 5 times 2 to 5 X increases in conversion rates and in the employee learning and development use case some of our

40:13 customers are seeing 60 percent increases in engagement rates and you know there's 300 percent faster ramp up in learning and 54 percent increase in confidence scores.

40:24 And lastly in the interview AI interviewing use case we're seeing 23 percent more completed interviews versus audio alone so you know every single customer use case we're seeing like really like tangible strong results.

40:40 That's incredible. I'm really excited to see some of these go into more the mainstream and get these use cases really really dialed in.

40:49A discovery-agent idea for LeanScale + wrap

40:49 I mean even for us at LeanScale one thing we have to do a lot of discovery for our engagements we have to learn what's going into the business.

40:57 There's we can pick up a lot with like research and connecting their systems and get a sense of what's going on.

41:03 But at the end of the day you still need to ask the people questions and that takes a lot of human time from our side.

41:10 So it's it's meetings or it's slack or having them fill out intakes or things like that.

41:15 And I'm even as we're talking about this I could totally see like a discovery or research agent process where it's easy for them they can schedule whenever they want.

41:25 And if you want to do it at midnight go ahead and and meet with it then and it's not dependent on our time and then it can save our team's time on hours and hours of discovery.

41:36 Definitely. Well you know if you become a customer we'll give you a nice podcast discount applies for any of the potential customers listening to this podcast the special Anthony discount.

41:45 I love it. We'll put the Anthony code in the description so they can use it.

41:50 Thank you. Thank you so much for going through everything such an inspiring and amazing story of where you've been and how you got to where you are today.

42:00 And I also love the passion outside of work for poker and then how you apply that to the game of business in the game and go to market and appreciate what you're doing in the world of AI and helping to push these things forward.

42:15 So thank you for being on the podcast and looking forward to what you do next.

42:19 Thank you Anthony. It's a pleasure being here.