Sep 7, 2026 · 1h 16m · news
Town vs Instinct vs GrokBot | Why the AI Assistant Market Is Not a Bubble
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this 20VC interview, Harry Stebbings and Town.com founder Jean-Denis Greze examine the competitive landscape of AI assistants, enterprise monetization models, and the paradox of building at machine speed while learning at human speed.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 14.2% of the talking time here. How this is scored →
Harry as informed peer 5.5 Guest teaching 4.7 Guest disagreement 2.7 Harry pushing back 4.0
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:02:00 Condemning Valley valuation inflating tactics
Jean-Denis forcefully rejects the premise of hype-driven fundraising announcements, calling out founders who raise small tranches at high valuations to falsely advertise inflated headline numbers as unethical.
Hardest push from Harry ▶ 25:11 Harry challenges Town's task sophisticationHarry directly dismisses Town's complexity compared to Instinct, stating that email tagging and pre-briefs are trivial tasks that should not require costly frontier models.
Biggest teaching moment ▶ 13:15 Autonomous data permissioning across silosJean-Denis educates Harry on the future of autonomous agent data negotiations, outlining how LLMs will replace traditional human compliance and security teams in permissioning sensitive enterprise data.
Harry holds his own ▶ 57:03 Harry pushes enterprise B2B over parent consumer tractionHarry leverages his venture experience to dismantle the attractiveness of B2C parent traction, contrasting the virality of suburban parent groups with landing enterprise accounts like Revolut with 7,000 seats.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Introducing Jean-Denis Greze and Town.com | 4 | 3 | 2 | 2 | Harry opens with an authentic, friendly endorsement of Town as a daily active user before bluntly characterizing Jean-Denis's previous startup as 'boring shit in finance.' Jean-Denis warmly accepts the critique and details the technical and market shift that prompted their pivot to an agentic email assistant. | |
| Big Tech Threats and Building Defensible Moats | 5 | 5 | 3 | 4 | Harry challenges Jean-Denis on the existential threat of frontier model providers and GrokBot eating Town's lunch. Jean-Denis reframes the moat debate as an early-stage luxury, arguing that real defensibility will come from agent-to-agent network effects rather than foundational models. | |
| Single-Agent Entry Points vs. Multi-Agent Ecosystems | 5 | 4 | 2 | 3 | Harry probes the architecture of human-agent interfaces, asking whether users will adopt multiple specialized agents or a single interface. Jean-Denis points to data privacy silos between personal and corporate contexts as the key determining factor. | |
| Trusting Autonomous Agents with Enterprise Data Silos | 4 | 7 | 2 | 2 | Jean-Denis educates Harry with a bold prediction: in five years humans will trust LLMs to autonomously negotiate and permission cross-silo data sharing. He draws upon information theory and enterprise sales dynamics to explain how removing human-mediated data filtering boosts efficiency. | |
| Managing AI Error Rates and Goal-Seeking Tendencies | 6 | 4 | 3 | 5 | Harry challenges the guest on error tolerance and quotes Jason Lemkin regarding dangerous agentic goal-seeking. Jean-Denis counters that smart humans make catastrophic errors all the time and frames the human role as an allocator of token budgets and high-level monitor. | |
| Model Routing, Persona Consistency, and Infrastructure Costs | 5 | 6 | 2 | 3 | Harry questions model routing and infrastructure dependencies. Jean-Denis provides deep technical nuances, explaining why routing is difficult in user-facing conversational layers due to Anthropics persona consistency, whereas coding layers can be routed flexibly. | |
| Allocating Engineering Resources Between Frontier and Open-Source Models | 7 | 5 | 4 | 7 | Harry directly confronts Jean-Denis, stating with 'the greatest of respects' that Town's tasks like email tagging are far simpler than Instinct's complex web scraping tasks and do not warrant expensive frontier models. Jean-Denis defends his prioritization, explaining that engineering hours are better spent expanding product value than prematurely optimizing COGS. | |
| Enterprise Virality Through Tinkerers and Underserved Teams | 5 | 4 | 1 | 2 | Harry asks for lessons in driving enterprise expansion. Jean-Denis highlights how identifying lone tinkerers who build reusable team skills creates viral adoption, particularly inside operational teams that AI vendors traditionally overlook. | |
| Delivering Immediate Delight Through Zero-Config Automation | 6 | 5 | 3 | 4 | Harry notes being pitched multiple European clones of Town and Instinct and asks how investors should evaluate them. Jean-Denis warns that local clones face brutal unit economics because they lack the war chest needed to keep pace with frontier R&D like OpenAI Codex. | |
| Machine Speed Building vs. Human Speed Learning | 4 | 6 | 2 | 2 | Jean-Denis reflects on the unprecedented velocity of the current market, noting that while software can now be built at machine speed, customer comprehension still happens at human speed. He describes the intense pressure of competing against nimble teams at Anthropic and Cursor. | |
| Differentiating Town from Competitors Instinct and GrokBot | 5 | 4 | 4 | 4 | Harry tries to bait Jean-Denis into naming his top three competitors using a 'Louis Theroux' questioning style. Jean-Denis flatly refuses to provide free marketing, contrasting Town's B2B monetization and multi-player focus against Instinct's subsidized consumer play. | |
| Dissecting Apple's Agent Strategy and Hardware Limits | 5 | 6 | 2 | 2 | Harry asks Jean-Denis to break down Apple's agent roadmap. Jean-Denis delivers a sharp critique of Apple's fundamental handicaps: a lack of cloud expertise and a rigid commitment to on-device privacy that leaves their local models months behind the frontier. | |
| Code Generation, Cyber Threats, and Industry Guardrails | 5 | 5 | 2 | 3 | Harry raises the looming danger of AI-driven cybersecurity breaches. Jean-Denis asserts that humans will never inspect every line of code again, comparing the developing regulatory and testing guardrails to early EPA environmental frameworks. | |
| Structuring Pricing Tiers and Curbing Token Waste | 6 | 4 | 2 | 4 | Harry inquires about customer success metrics and tier profitability, bringing up examples of power users consuming thousands in tokens. Jean-Denis explains why token maximization is dangerous and describes introducing proactive alerts to prevent customers from wasting tokens on rogue routines. | |
| Prioritizing High Net Revenue Retention in B2B Over B2C Scale | 6 | 5 | 3 | 4 | Harry presents a forced choice between 100M consumers at $20/month versus 1M enterprise users at $100/month. Jean-Denis argues firmly for B2B accounts, emphasizing that business workflows offer unbounded net revenue expansion compared to capped personal use cases. | |
| Voice Capabilities, ElevenLabs, and Supplier Margin Squeeze | 6 | 5 | 3 | 6 | Harry pushes on ElevenLabs' rumored $22B valuation and asks if high voice inference costs will crush Town's gross margins. Jean-Denis acknowledges the strategic danger of paying high margins to upstream suppliers while competing with them directly. | |
| The Onboarding Trade-Off: Requiring Calendar and Email Access | 7 | 5 | 5 | 7 | Harry directly challenges Town's core premise, asking whether requiring calendar and email access is truly insightful since most users resist giving away inbox permissions. Jean-Denis holds firm, revealing they accept a deliberate 30% onboarding churn to guarantee immediate automated value for the rest. | |
| Navigating Unexpected B2C Product-Market Fit with Families | 7 | 3 | 2 | 6 | Jean-Denis shares an internal debate over unexpected traction among parents. Harry forcefully pushes back, mocking the expansion potential of 'people in Sonoma with kids under five' compared to enterprise accounts like Revolut with thousands of seats. | |
| Valuation Rumors, PR Tactics, and Fundraising Ethics | 7 | 3 | 5 | 6 | Harry presses on Town's revenue figures and rumored billion-dollar valuation, offering PR advice on separating milestones. Jean-Denis declines to disclose numbers and delivers an impassioned critique of deceptive Silicon Valley fundraising PR tactics. | |
| Quickfire Insights: Angel Bets, $100B Scale, and Board Needs | 5 | 4 | 3 | 4 | In the quickfire round, Jean-Denis outlines the math for a $100B company ($700+ ARR across 10M paying users) and mentions wanting a late-stage CFO on his board. Harry reacts with skepticism, remarking that seeking an operational CFO seems premature. | |
| Resisting Infinite Cash Burn to Validate True Value | 6 | 5 | 3 | 5 | Harry challenges Jean-Denis on why he doesn't completely subsidize user acquisition and 'burn the boats' to capture land. Jean-Denis argues that forcing enterprise customers to pay immediately is the only way to validate genuine product value and avoid artificial token usage. | |
| Rethinking Tech Hiring: Skipping Traditional Technical Interviews | 6 | 5 | 2 | 3 | Jean-Denis explains why Town bypasses coding interviews for trusted referrals. Harry connects this to venture frameworks from Brian Singerman and Marc Benioff's developer tooling ratios, leading into Jean-Denis's analysis of developer ROI versus compute costs. |
Statements from this episode (60)
Greze: Town pivoted after spending a year failing at AI tax prep
“We've spent a year building an AI tax company, like business tax prep with AI, and we just, we got to some product market fit, but not enough where it was going to be a success. Like that's, you know, there's a thing, like the truth is we failed at building a …”
Greze: Town's two-week AI prototype found product-market fit almost immediately
“We built a quick prototype in a couple of weeks and it had product market fit almost immediately.”
Greze: Building an AI assistant is a top-three priority at Apple and Google
“You want to know, the truth is I know what I'm building is a top three priority at Google and Apple, like in the next 12 months. Not a top 10 priority, like a top three priority.”
Greze: GrokBot is a power-user product, not a mainstream one
“Grokbot's really cool. It's awesome, but it's a power user product. It's not a mainstream product.”
Greze: Big tech incumbents will copy rather than innovate mainstream AI products
“The big players are not, they're not going to innovate their way there. They will copy their way there, but they have to copy someone who's been successful in the first place at building a mainstream product.”
Greze: Multiplayer AI remains unsolved and will become the category's moat
“I think no one has figured out multi-user, multi-player AI today. Like, I think that's the thing that will be the moat.”
Greze: Meta will launch a personal AI assistant inside WhatsApp
“They're going to have a personal assistant that's going to come in WhatsApp. I don't know if they're launching it in a day or in three months, but it's coming.”
Greze: Users will rely on one to three digital entry points, not dozens
“Each human will have one, two, maybe three entry points into the digital space, because I don't think you'll want to be like, oh, I'm doing sales. Let me use the Salesforce agent. Oh, I'm doing project management. Let me use the linear agent. Oh, I'm doing thi…”
Greze: Enterprise and personal AI data must always be separated at the data layer
“From a privacy perspective and from a, like, where your data lives, I think you're going to always want to separate on the data layer your personal and your work data.”
Greze predicts users will trust AI agents to autonomously share personal data
“I think you'll trust your agent to decide what data to share with other people without you intervening in five years.”
Greze: LLMs will replace human security teams in classifying and gating data
“As opposed to having humans in your compliance and security team over time, label data and decide what goes where and what can be accessed. I think we'll just start to trust LLMs to do that.”
Greze: LLMs will make far fewer workplace mistakes than humans pretty quickly
“I think the LLMs will make many fewer of these mistakes than humans pretty quickly.”
Greze: LLMs fundamentally turn energy into GDP and revenue
“One way to think about LLMs, right, is they're just, they turn energy into, like, GDP, right, or into revenue.”
Greze: Humans must set goals, token budgets, and monitor AI agent actions
“I think it is the human's responsibility to first allocate the resources. That means to say how many tokens are we willing to spend to try to get a goal to set the goal right as well. So you set the goal and the budget and then to monitor, right? Like the over…”
Greze: Anthropic spends significant effort keeping model personalities consistent
“One of the problems when you do model routing is there's some companies like Anthropix spends a lot of time. I know we make fun of them online, but they spend a lot of time actually making sure all of their model families roughly don't change too much in terms…”
Greze: User-facing persona limits AI model routing unlike backend reasoning
“The way I think about our stack is there is the part of the stack that deals with the user interface, like the feel and the personality. And there it is harder for me to just route Wildly, because I need consistency of the experience that is sometimes hard to …”
Greze: AI startups subsidize inference costs betting on 18-24 month cost declines
“The answer there for every startup that I know outside of a very few. Is, yeah, it's, well, we're hoping the price, we're hoping the cost curve makes it efficient in 18 to 24 months, right? In the meantime, you're subsidizing in part, right? Because that's wha…”
Greze: Open-weight models already perform routine human-level tasks really well
“Scheduling movies, working with one's calendar, answering emails that have been answered before, doing research on competitors on a daily basis. Like, all these kinds of things, I think, are trending pretty far from the frontier, right? And you can already use…”
Greze: Moving to open-weight models improves COGS without providing product advantage
“I can improve our cogs by moving to open weight, but it doesn't give me much product advantage. It doesn't make my product work better.”
Greze: Enterprise AI adoption accelerates when a team has an internal tinkerer
“If the company has one person who is a tinkerer and who starts to build things like team skills, team integrations, team routines, that is like building, those are all building blocks on town that everyone on the team gets for free. Then we'd send to see a lot…”
Greze: Town drives early growth via underserved roles like HR and EAs
“There are other functions, like executive assistants, chiefs of staff HR team members, finance, like, some more junior finance team members that don't, there's, don't have that much AI in their day-to-day. They, like, really don't. And a lot of their workflows…”
Greze: Very low time to value is the only reason Town works
“I think the only reason our product works today, honestly, is we have very low time to value for a single user.”
Stebbings has been pitched 4 to 5 European clones of Town and Instinct
“I've been pitched four to five European Towns or European instincts. I mean, literally four to five separate ones.”
Greze: Codex has roughly 100 people working on it
“Codex is like a hundred people making that thing better”
Greze: Silicon Valley hiring is no harder now than competing with Stripe at Plaid
“I don't find it crazy, honestly. Like I think when I was at Plaid and we were competing with talent for like Stripe, that felt like no harder than what I'm doing now.”
Greze: Startups can build at machine speed but only learn at human speed
“You can build now at the speed of machines, but you can only learn at the speed of humans.”
Greze: AI market is a blue ocean because users treat ChatGPT as a Google enhancer
“I mean, you know, look, it's a blue ocean market, right? You gotta understand. Like I never, when we go to most customers, they've not heard of anything. It's blue ocean because people, people are using ChatGPT as a Google enhancer. Like that's the market.”
Greze: Town monetizes enterprise teams whereas Instinct relies on subsidized free acquisition
“I think what we do and generate, we generate revenue, right? From companies, right, that are using us for work with network effects around multiple team members working on it. Like, their product doesn't do any of that. Maybe that is part of their strategy. I …”
Greze: GrokBot faces permanent enterprise resistance due to its brand
“I think some people just won't want to touch it because of brand. And it's like, that's just inevitable. And it's, you know, that's just a thing that they're going to have to deal with forever.”
Greze: AI adoption by users on X is not real product-market fit
“I think a person on X that uses these products is not actually product market fit. Meaning that those are not the mainstream users and you have to keep that in mind.”
Greze: Apple lacks cloud DNA and does not know how to do cloud
“One, they're not a cloud company. They're just not, it's just not their DNA. They don't know how to do cloud.”
Greze: On-device phone AI models are slower and dumber than frontier models
“Like, local models on the phone, it's amazing, but they're just not, they're just, it's just slower and dumber, right, than what's at the frontier.”
Greze: Apple's new Siri will lag Town and GrokBot by nine months in capability
“But I think it's going to be good, but it's going to feel not nearly as powerful as Town or GrokBot. Like it's not even going to be there in terms of its capabilities, but it'll be on your phone. It'll be convenient. You'll be able to like enable more data wit…”
Greze: Humans will never again read every line of code
“We've passed the point where humans will read every line of code. That is never happening again.”
Greze: A successful user is simply someone who pays monthly
“Oh well, I just define them as someone who pays me every month. If they keep paying me, no, I'm serious. I'm serious, right? If they keep paying me, I've done my job, right?”
Grèze: Measuring AI customer success by token consumption causes churn
“It's a dangerous way to think about it, because if you think about it as like, they use more tokens every month that's successful. What if they're using the tokens in a way where the ROI is less clear to them? Meaning they don't realize that they're using toke…”
Stebbings: Jason Lemkin burns $15k in tokens on a $299 Anthropic plan
“My friend Jason Lampkin, he obviously pays for, like, Anthropic Pro, whatever it is, two 99, and he spends about 15,000 dollars of tokens. He is the worst customer for Anthropic, but he's on that, like, pro max individual plan.”
Grèze: Town's $15 tier has the worst unit economics and is subsidized
“The 15 dollar plan is a really good deal. I would say for users, it's mostly, I would say it's like a, we use it as a way for people to use the product enough that they realize they should pay 49 where the product is really powerful. So 15 has the worst, the 1…”
Grèze: Town's $99 tier is probably its most profitable overall
“And then I would say probably the 99 dollar plan is the most profitable overall because it's like a power user, but it's not a power user that's like trying to like, You know, spend on limited numbers of spend, but we also have usage based pricing, right?”
Greze: B2B AI enables revenue generation whereas consumer AI only saves time
“And when I do those things, I save time and time is worth money. But on the business side, when I create something that generates value for the business, They make more money and then they want more of that thing.”
Greze: Conversational voice will be incredible in two to three years
“I mean, voice. Voice is so obvious. It's, I mean, it's happening right now, but it's not voice like you just speak to it. I just mean conversational.”
Greze: ElevenLabs risks voice quality topping out against open-weight models
“What I don't know is if it tops out, and that's, I think, the risk for something like 11 Labs, meaning, like, we just get voices good enough, and then you can get it. I can put, you know, open weight models on base 10 and get it, but it just doesn't feel like …”
Greze: AI wrappers face broken economics paying suppliers 70% margins to compete
“And the problem with the frontier is, I have zero pricing power at the frontier. And I mean, I think this is what happened to Cursor, right, at the end. It's like, you can have huge market share and customers love you and everything, but if you're paying your …”
Greze: Over 15% of users who try Town convert to paid
“The payment rate for us on acquisition is like more than 15% of users who try the product end up paying for it, which is extremely high for PLG, because the value delivered relative to what they were getting out of ChatGPT is just huge.”
Greze: ChatGPT's automated suggestions are plain bad
“I think their suggestions are, I think their suggestions are just like plain bad, to be honest, but they didn't even have suggestions until a few months ago.”
Greze: 30% of users churn immediately when required to connect email and calendar
“30%. Right off the bat.”
Greze: Town has tremendous product-market fit among families
“Parents and families, there's like tremendous product market fit for town there.”
Stebbings: Founders should separate fundraise and revenue announcements for PR
“Too many times I see people like combine a fundraise with a revenue milestone. Do not do that. Those are two separate PR moments that can be made into two big moments, not one.”
Stebbings: Instinct raised at a $2.5 billion valuation without monetization
“Instinct raised at two and a half billion dollars with no monetization.”
Greze: Silicon Valley startups routinely inflate headline valuations via blended rounds
“I have seen deals where it's like, I invested like 200 and then the announcement is at 500. And then what you learn is that like, you know, they raised sixty five million and like five millions at 500 and the other 60 is like 200 or 300, right? There's a whole…”
Greze: Misleading headline valuations on fundraising rounds is unethical toward employees
“Personally, I don't think it's ethical. I don't think it's ethical towards employees. You know, most of all, if you're not, I mean, maybe when you hire someone, you tell them for sure. Because it's like the dilution wasn't that number one. It's not where most …”
Greze: Town can hit $100B valuation with 10M paying users
“I think if we can get about ten million people paying for the product, we can get to that.”
Greze: Town generates over $700 per user annually
“I mean, our, yeah, we make over 707 hundred dollars per year per user today.”
Greze: One beta user racked up $26,000 in compute over five months
“There were users who were spending a shit you not like 2000 dollars of compute a month, 4000 dollars of compute a month. There's someone on the platform who'd spent in five months was like 26,000 dollars.”
Greze: Enterprise customers dislike free software due to unpredictable future costs
“But what I've learned is like on the business side, They also don't like it if you don't charge them because they don't know how much it's going to cost one day. They want to know how much it's going to cost one day. You can't sell to a 500 person company and …”
Greze: The AI assistant economy will be paid, not ad-supported
“So I'm a big believer that actually the economy around assistance will be paid for.”
Greze: Town skips core role evaluation for vouched-for engineering candidates
“A person that I trust that's great on my team, like great on my team. They tell me this other person is like one of the best people that I ever worked with. And then I'm going to make that person like spend eight hours doing stupid whiteboard interview or like…”
Stebbings: Brian Singerman automatically invests when a fund manager is all in
“Brian Singerman, who invests in funds and then invests in the companies beneath those funds, has a rule that if the manager's like, balls to the wall, I am all in on this company, he'll automatically write the check.”
Greze: Town spends at least $75K per engineer on developer tools
“I mean, the run rate's at least 75 K per employee, per engineer.”
Greze: AI encourages hiring more engineers by boosting their marginal revenue
“AI means that suddenly that engineer can generate more than they could have before. So maybe before they could only generate a 150 K of revenue. Maybe now they can generate two 50 K of revenue. So suddenly AI makes you hire the incremental person, right? One m…”