Alex Rampell

1. Key Themes

Killing a "Good" Business Because Tax Is a Discontinuous, Non-Viral Market

Greze walked away from a SMB tax-automation company with real traction because the market's structure capped its scale. The seasonal, once-a-year interaction meant no word of mouth, no DAU, and no way to outrun incumbents who could bolt AI onto existing customers. He frames the decision as a function of venture scale plus incumbent catch-up speed, not of whether the product worked.

"Pre-AI you would have been like, 'That makes a ton of sense.' But post-AI this was not going to be a venture scale company." — Jean-Denis Greze [00:04:13]
"It felt for tax like we wouldn't be able to get to hundreds of millions of revenue before our competitors, the incumbents could, because you can only sell tax once a year." — Jean-Denis Greze [00:05:50]
"It also doesn't have a ton of word-of-mouth, because it's rare for people to brag about their tax prep situation." — Jean-Denis Greze [00:05:57]

The Core Problem in AI Adoption Is "Tinkerers," Not Intelligence

Greze argues that model capability is already sufficient for most business problems. Diffusion is blocked because only a minority of people will experiment with a text box. The winning product proactively shows value rather than teaching users to prompt.

"The reason why it's not diffusing is because even though in Claude or in ChatGPT, there has existed for the last 12 months features that at a lot of small businesses you could use to save you so much more time, but instead what you're doing is you're using them as a replacement, as a like a better Google search." — Jean-Denis Greze [00:46:40]
"You can't teach someone. You have to just show the value prop to someone and do it for them." — Jean-Denis Greze [00:47:16]
"You don't need more intelligence, the intelligence is there. It's just, how do you deploy it effectively at scale?" — Jean-Denis Greze [00:48:24]

Token Allocation: AI Makes Long-Tail Work Economically Viable

Greze describes Town as a "token allocation machine" that surfaces places to spend cheap tokens on tasks no human or EA would ever have done. A restaurant owner example shows the asymmetric payoff: pennies most months, hundreds or thousands of dollars in the month it catches a price hike.

"I think of our product as a token allocation machine. We're trying to give options to the human about where they could apply tokens." — Jean-Denis Greze [00:23:19]
"If it finds nothing 11 out of 12 months that's worth doing, it's going to cost you like a dollar a month to run that thing. But the interesting thing is the 1 month out of a year where it finds something, it's going to find like hundreds, if not thousands of dollars for you." — Jean-Denis Greze [00:25:24]
Rampell ties it to Jevons paradox: "it turns out, if you could actually produce it via silicon, the market's quite large." [00:22:12]

Horizontal-First, Then Let Usage Data Define Verticals (Without Reading User Data)

Town stays horizontal and uses aggregate usage signals to find power users, interviews them (increasingly AI-run interviews), and hand-carves "routines" templates per profession. This is a semi-manual flywheel that gets automated as models improve.

"We look at aggregate metrics of which... are our power users on the platform today. And then we go talk to them." — Jean-Denis Greze [00:30:59]
"We'll offer 80/20 very customized software for people, and people aren't just one role... you run finance for your restaurant, too... And also you have a vacation rental." — Jean-Denis Greze [00:32:25]
"You give them a free Amazon card, and you're doing an AI recorded interview of them... you're no longer operating at the speed of humans, you're operating at the speed of AI." — Jean-Denis Greze [00:31:54]

Neutrality and Data Ownership as the Wedge Against the Labs

Town's pitch to mid-market firms is model-agnosticism, ownership of the context layer with one-day switching costs, and no training on customer data. Greze deliberately lowers his own switching costs as a selling point.

"We're a neutral app. We will use whatever supplier we think is the best for you in the moment." — Jean-Denis Greze [00:36:34]
"There are companies that picked OpenAI like 2 years ago, and then Anthropic was better at coding for a 12-month period, and those companies had to use both." — Jean-Denis Greze [00:37:17]
"If there was ever a better solution than Town, switching to that other solution would take, you know, a day." — Jean-Denis Greze [00:38:22]
"If you look at anyone who's deploying FDEs to build software for you, they actually... are trying to make it more difficult for you to switch." — Jean-Denis Greze [00:38:22]

Consumer-Grade Product First, Enterprise Later (the iPhone/BlackBerry Pattern)

Drawing on his Dropbox experience, Greze says you can bolt enterprise compliance onto a great product, but cannot start from compliance and reach delight. Rampell links this to CEOs getting iPhones for Christmas and killing the BlackBerry.

"It is almost impossible to go from an experience that starts from the compliant IT enterprise requirements and then figures out how to make the product experience incredible." — Jean-Denis Greze [00:14:06]
"Whoever has an enterprise experience that feels like a consumer experience like they're going to be set up to win." — Jean-Denis Greze [00:14:06]
He explicitly fears becoming "Good" (the ugly enterprise iPhone email client): "I don't want to be Good, right? Because Good eventually is not a good business." [00:17:41]

Agents Break the Attention-for-Content Covenant and Devalue Supply Aggregation

Agent traffic is changing the economics of the internet. Amazon's adversarial posture is understandable given its ad business, but Greze argues supply aggregation loses value when human attention can be automated away, and that neutral agents paid only by the user's wallet are the logical endpoint.

"I just think supply aggregation is a lot less valuable. It used to be very valuable because human attention was limited, and now... human attention can be automated away." — Jean-Denis Greze [01:01:15]
"I'm a firm believer in... you want a neutral agent that is yours, that you have no de- like the only source of money for the agent is your wallet as the user." — Jean-Denis Greze [00:54:26]
Rampell: "The covenant is like, I, Yahoo!, give you, Alex, free stock quote if you give me your eyeball. And that covenant has been broken." [00:52:59]

Friction You Enjoy vs. Friction You Don't

Greze argues agent builders over-index on efficiency. Humans enjoy some friction (browsing Soho, discovering non-Lego toys through ads), which preserves a role for aggregators, who must build a distinct experience where ads are wanted.

"Sometimes when we think about agents too much, we think too much about reducing friction, and toil... but humans don't always optimize, you know, for efficiency." — Jean-Denis Greze [00:56:02]
"I think there may be a bigger than Amazon company, like a headless Amazon, could be a huge huge business." — Jean-Denis Greze [00:57:05]

Trust Agents to Mediate Internal Information at the Speed of AI

Town's Teams product lets agents query other employees' agents and decide what to share, replacing slow human routing in Slack.

"We've decided that for that kind of a question, we will trust AI to mediate the security and privacy of that request within a company." — Jean-Denis Greze [00:41:05]
Rampell: "Humans are slow routers, that's right." [00:41:04]

2. Contrarian Perspectives

The Unit of Value in Software Moves From the App to the Computer

Greze says the app layer itself is being abstracted away, which undermines most SaaS valuation logic and feature-based differentiation.

"I think the unit of value in software has changed from the app to the computer... we're abstracting away the app layer totally, and I think the value will drop in the app layer." — Jean-Denis Greze [00:29:07]
"It's going to be more and more difficult to sustain that kind of an advantage" of feature leads. — Jean-Denis Greze [00:29:07]

The Labs' Forward-Deployed-Engineer Strategy Gets Worse as Models Improve

Most observers see FDEs as the winning enterprise AI go-to-market. Greze bets it is a transitional artifact that also functions as lock-in.

"If you're an enterprise that's willing to spend like $5 million a year paying OpenAI for FDEs, like it looks really good right now, but I think over time, as the model gets better, the FDE approach actually gets much worse." — Jean-Denis Greze [00:33:00]
"Let other people fight in the enterprise for this much more like handcrafted software." — Jean-Denis Greze [00:33:59]

A Small, Unsexy Horizontal Product Can Sit in the "Goldilocks Zone" the Labs Ignore (For Now)

Rampell argues coding fell on the wrong side of relevance, while horizontal work automation is not yet a must-win market for the labs. Greze counters that it is a "when, not if" problem, and the window is perhaps 12 months.

Rampell: "If you were a coding agent right now... Anthropic's adding so much. It's so relevant for them. They cannot ignore it." [00:42:20]
Greze: "It's a question of like, when, not if... in the current environment, it's not going to move the needle at those companies, but I think in 12 months it will." [00:44:10]

Don't Block Agents; Build the Counter-Agent

Conventional wisdom says defend your platform from agents. Greze says Amazon's best countermove is a Muse-like personal agent of its own with exclusive shopping, bundled in Prime, and that a "headless Amazon" could be enormous.

"The countermove for them honestly is they should have a personal agent... It's not to block Muse, it's to have a Muse." — Jean-Denis Greze [00:59:15]
"Amazon might lose a huge chunk of their advertising business trying to fight the future." — Jean-Denis Greze [00:57:36]

Agents Kill Price-Discrimination "Breakage" Models

Agents invert the time-vs-money segmentation that coupons, rewards programs, and card benefits depend on.

Rampell: "People that have more money than time... what if you now can give everybody who has more money than time, give them all of the tools, and the agent does it?" [01:00:17]
Greze: "We're all going to abuse all of our Chase and Amex benefits." [01:01:06] Rampell: "The breakage model is dead." [01:01:06]

3. Companies Identified

Town

Description: AI assistant ("Townie") for email, calendar, and everyday business workflows, with Teams features; a16z-backed.
Why mentioned: The subject company; reportedly changed Ben Horowitz's life; strong PLG growth, including unplanned traction in Japan.
Quotes: "I get pinged by Ben Horowitz, about like, 'I found this amazing tool. It's called Town. It changed my life.'" — Alex Rampell [00:03:17]. "We're big in Japan. That wasn't intentful." — Alex Rampell [00:45:26]

Plaid

Description: Fintech data connectivity company; Greze was CTO; a16z investor.
Why mentioned: Origin of the Rampell–Greze relationship.
Quotes: "We worked together when you were the CTO of Plaid, and I had led our investment in Plaid here." — Alex Rampell [00:02:06]

Dropbox

Description: File storage and collaboration company.
Why mentioned: Greze and his co-founder worked there; source of the consumer-first-then-enterprise lesson.
Quotes: "My co-founder and I worked at Dropbox, and I think one of the big learnings for us from Dropbox is you can always take a world-class product experience and figure out how it works for the enterprise." — Jean-Denis Greze [00:14:06]

Anthropic

Description: AI lab behind Claude.
Why mentioned: Went all in on coding while OpenAI spread across 50 bets; described as supplier and potential competitor; added "so much" so fast that it cannot ignore coding.
Quotes: "OpenAI had like 50 bets, and then Anthropic went all in on code, and that did really, really well." — Alex Rampell [00:43:26]

OpenAI

Description: AI lab behind ChatGPT.
Why mentioned: Supplier and potential competitor; building vertical companies that compete with startups; Rampell used ChatGPT to rewrite a Harvard professor's dissertation.
Quotes: "If you look in OpenAI, they're building vertical companies that are competing with that." — Jean-Denis Greze [00:42:01]

Rogo and Harvey

Description: Vertical AI companies (finance and legal).
Why mentioned: Cited as cautionary examples of labs moving up the stack.
Quotes: "Look at what's happening to Rogo. What's happening to Harvey, right?" — Jean-Denis Greze [00:42:01]

Walmart

Description: Retailer without a large advertising business.
Why mentioned: Potentially advantaged by agent-friendly posture.
Quotes: "Amazon's loss should be Walmart's gain, right? ... they don't have the advertising business, and so their incentives are going to be different." — Jean-Denis Greze [00:53:45]

Amazon

Description: E-commerce and advertising giant.
Why mentioned: Adversarial stance toward agents; ~$70–80B advertising business at risk; the "headless Amazon" idea.
Quotes: "Amazon has obviously adopted a somewhat adversarial approach, which makes sense because they're an advertising company." — Alex Rampell [00:52:30]

Meta (Muse and Instagram)

Description: Social platform company with a free personal agent, Muse.
Why mentioned: Example of divorcing monetization from the agent experience: free agent feeds data to better ad targeting and a wallet on Instagram.
Quotes: "Maybe Muse just gives them data that allows to target you better on Instagram." — Jean-Denis Greze [00:58:05]

CamelCamelCamel

Description: Amazon price-tracking, affiliate-funded website, around 18 years old.
Why mentioned: Evidence that price-tracking agents are already the "present future"; ~20 million users in America.
Quotes: "20 million people use this website... it's an affiliate website, and this has been around for 18 years." — Alex Rampell [00:59:50]

Good Technology

Description: Enterprise-ready email client for the iPhone.
Why mentioned: The cautionary tale of an enterprise-compliant but ugly product.
Quotes: "It was like the ugliest thing in the history of the world." — Alex Rampell [00:17:38]

BlackBerry / Microsoft / Nokia

Description: Incumbent mobile players.
Why mentioned: Analogy for enterprise leaders adopting consumer AI.
Quotes: "Steve Ballmer's like, 'Nobody's going to buy an $800 phone.'" — Alex Rampell [00:16:19]

Cisco, Akamai, Amazon, eBay, Yahoo

Description: 1990s infrastructure and application companies.
Why mentioned: Historical analogy for infrastructure-versus-application layering, where infrastructure vendors stayed out of applications.
Quotes: "Cisco and Akamai were infrastructure companies in the '90s, and then Amazon and eBay were application companies." — Alex Rampell [00:34:32]

Yahoo! Finance, E*TRADE, NASDAQ

Description: Early ad-supported finance portal, advertiser, and data source.
Why mentioned: Illustration of the free-content-for-eyeballs covenant.
Quotes: "I go to Yahoo! Finance to look up a stock quote... my eyeballs then hit an E*TRADE ad." — Alex Rampell [00:52:30]

Canva

Description: Design software.
Why mentioned: Example of agents using existing user subscriptions via computer use.
Quotes: "I can do it using your Canva subscription, right? Cuz you're already paying for it." — Jean-Denis Greze [00:51:18]

Slack, Microsoft Teams, LinkedIn, Gmail

Description: Workplace and communication tools.
Why mentioned: Connectors and the slow human-routing baseline.
Quotes: "You would go in Slack or Microsoft Teams and be like, 'Does anyone know someone at that team?'" — Jean-Denis Greze [00:40:38]

GEICO, Amex, Chase

Description: Insurance and card issuers.
Why mentioned: Pitch-framing analogy and breakage-reliant benefit programs.
Quotes: "Amex premium membership, it's the same way." — Jean-Denis Greze [01:01:02]

Google

Description: Search and AI company.
Why mentioned: One of the labs that could decide to build in Town's space.
Quotes: "Someone at OpenAI, Anthropic, at Google could wake up and decide to do that." — Jean-Denis Greze [00:35:38]

Claude / Opus 4.5

Description: Anthropic's model, whose agentic leap helped Town reach product-market fit.
Why mentioned: Timing unlock.
Quotes: "We did it at the time that Opus 45 was happening and agents... were able to do more and more autonomous work." — Jean-Denis Greze [00:11:15]

4. People Identified

Jean-Denis Greze

Description: Founder and CEO of Town; former CTO of Plaid; ex-Dropbox.
Why mentioned: Made a rare, disciplined pivot away from a ~$1M-ARR tax business; coherent thesis on neutrality, PLG, and agents.
Quotes: "In the moment, it felt like failure... but I think, what are your core motivations for building a business?" — Jean-Denis Greze [00:07:19]

Description: General Partner at a16z; led the Plaid investment; investor and power user in Town.
Why mentioned: Host; contributes the Jevons paradox, coupon/breakage, and Goldilocks-zone framings.
Quotes: "I'm both an investor and a power user of Town." — Alex Rampell [00:34:00]

Ben Horowitz

Description: Co-founder of a16z.
Why mentioned: Early power user who discovered Town and said it changed his life, which was the first signal to Rampell.
Quotes: "I get pinged by Ben Horowitz... 'It changed my life. It's so great.'" — Alex Rampell [00:03:17]

Steve Ballmer

Description: Former Microsoft CEO.
Why mentioned: Dismissed the iPhone; an example of incumbent denial.
Quotes: "Steve Ballmer's like, 'Nobody's going to buy an $800 phone.'" — Alex Rampell [00:16:19]

James Watt

Description: Inventor of the more efficient steam engine.
Why mentioned: Origin of the Jevons paradox analogy.
Quotes: "The original like James Watt made a more efficient steam engine." — Alex Rampell [00:23:19]

Town's co-founder (unnamed)

Description: Greze's co-founder, also ex-Dropbox.
Why mentioned: Joint decision-maker on the pivot.
Quotes: "My co-founder and I, we went for a lot of walks with a lot of coffee for a week." — Jean-Denis Greze [00:07:15]

Greze's edtech founder friend (unnamed)

Description: Founder of a successful edtech company.
Why mentioned: Corroborated the seasonality-limits-scale lesson.
Quotes: "You can only sell an edtech twice a year... 'That really limited my ability to hyperscale.'" — Jean-Denis Greze [00:06:18]

Justin (a16z colleague, unnamed surname)

Description: Co-author with Rampell of a post on observing the present to predict the future.
Why mentioned: CamelCamelCamel thesis.
Quotes: "Justin and I wrote a post about this, where the best way to predict the future is to observe the present." — Alex Rampell [01:00:16]

Harvard professor (unnamed)

Description: Senior Harvard administrator skeptical after trying early AI.
Why mentioned: Illustrates ossified mental models of AI.
Quotes: "I had ChatGPT rewrite her dissertation... She was like, 'This is amazing. What AI is this?'" — Alex Rampell [00:49:28]

5. Operating Insights

Use a Natural Decision Window as a Forcing Function

Because tax is annual, the team had roughly a month (around October 15th) in which they could still change direction. Greze credits the deadline with making a hard call efficient. Founders in non-cyclical businesses can impose their own windows to avoid endless "one more idea" drift.

"We had a short window, right, where we could change the direction potentially, and once the window was gone... a year from now we won't have the capital to make the change." — Jean-Denis Greze [00:10:15]

Build a Customer-Driven Template Library with Power-User Interviews

Find power users from aggregate metrics, interview about 10 per segment, extract the five common automations, and hand-build templates that seed onboarding. Then automate the interview and recruiting loop with AI-recorded interviews and gift-card incentives.

"We might talk to 10 power users that run restaurants. And what we'll see when we talk to that, there's five things in common that they all do." — Jean-Denis Greze [00:30:59]

Define "Stickiness" as One Custom Integration, Then Productize It Away

Retention correlates with connecting email, calendar, and one system of record, then building one custom thing that was magical. The goal is to deliver that value automatically at the moment of connection.

"I'm trying to just get rid of the custom integration. I'm going to get each person, which is as soon as you connect your system of record, can I deliver value to you in a way that feels very you." — Jean-Denis Greze [00:47:50]

Stage Enterprise Entry: Email and Calendar First, Systems of Record After "Earning the Right"

Win a few users per company with low-permission connections, then earn expansion to CRM and IT-managed tools. Meanwhile, build the CFO/CISO scaffolding (audit logs, privilege controls, token cost visibility) in parallel.

"Have I earned the right to ask you to connect your CRM?" — Jean-Denis Greze [00:19:00] (journey described at [00:17:28])
"You need the cost control for the CFO so that they understand that if their whole team operates on Town, they're understanding where money is being spent and that tokens are being used efficiently." — Jean-Denis Greze [00:19:00]

Suggest Concrete, Cheap-to-Try Actions, Not Capabilities

Don't market "computer use"; show a 25¢ outcome (posters in the user's Canva style) and then offer to repeat it automatically, and then extend it (video assets) as models improve.

"It needs to be much more concrete, like, 'Hey, like I can make posters for this event that like look just like the ones that you've been making in Canva.'" — Jean-Denis Greze [00:50:50]

6. Overlooked Insights

"Truly Owning" Context as a Third Business Asset Class, and Selling Low Switching Costs as the Moat

In one pass, Greze names a new category of corporate asset (the accumulated AI context about your business, alongside physical and human assets) and positions Town to be deliberately replaceable. If context becomes a core asset on the balance sheet of intelligence-driven firms, whoever lets customers own and port it gains trust, while FDE-led vendors who engineer lock-in face a growing procurement objection.

"There is the set of assets which is all the things that AI knows about your business to help you run your business effectively. And I think you should own that entire bucket." — Jean-Denis Greze [00:38:22]
"The risk that I'm taking, is that... the switching costs for my customers is easier, right? But that's my selling point." — Jean-Denis Greze [00:38:59]

The Free-Agent-Plus-Wallet Flywheel: Monetization Divorced from the Agent

Mentioned almost in passing, the Muse/Instagram example shows how a free personal agent can improve ad targeting elsewhere and then convert purchases through a wallet the agent has built, which means agent monetization may never appear as ads or fees inside the agent. That is the real threat to aggregator and marketplace economics, and the reason a neutral, user-paid agent is a differentiated position.

"You've divorced the monetization from the agent experience for this personal use case... Having just been the aggregator of supply just may not be a good business model in the world of agents." — Jean-Denis Greze [00:58:36]