At a glance
WHAT IT’S REALLY ABOUT
Why Town ditched SMB tax to build neutral AI work agents
- Jean-Denis Greze explains why he abandoned a technically successful SMB tax automation product: annual purchase cycles, weak word-of-mouth, and incumbent distribution made it unlikely to scale to venture outcomes.
- Town was born from customer pain and a market gap: a high-quality agent anchored in email and calendar that improves meeting prep, follow-ups, and everyday knowledge-work workflows.
- The company chose a product-led, consumer-grade experience first, arguing it’s easier to add enterprise controls (security, audit logs, cost governance) later than to retrofit usability into an enterprise-first product.
- Town’s strategy is to be horizontal yet deeply personalized, using observed power-user patterns to recommend role- and industry-specific automations while letting vertical opportunities emerge from usage.
- Greze argues AI adoption is under-penetrated outside Silicon Valley because most tools require tinkering and imagination; the winning platforms will “show” value with integrated workflows rather than rely on a text box interface.
IDEAS WORTH REMEMBERING
5 ideasIn AI-era markets, a working $1M business can still be the wrong company to build.
Greze built an AI-driven SMB tax product that worked technically and was approaching ~$1M ARR, but tax is purchased annually, has weak word-of-mouth, and gives incumbents time to copy while leveraging their distribution—making “venture scale” unlikely.
Email+calendar is a massive wedge because it’s cross-industry, global, and immediately useful.
Town emerged from a simple observation: email and calendar are universal workflow hubs, yet no one had built a truly great agent around meeting prep, follow-ups, and routine knowledge-work tasks; improving model agentic capability made PMF arrive quickly.
Start with the user experience; enterprise-readiness should follow, not lead.
They intentionally prioritized a consumer-grade UX and self-serve adoption (SMB/mid-market) rather than starting from enterprise compliance demands, believing it’s easier to add enterprise features later than to “consumerize” an IT-first product.
As tokens get cheaper, the surface area of economically-rational automation explodes.
Greze frames Town as a “token-allocation machine” that continuously suggests high-ROI automations as inference costs fall—enabling tasks no human EA would do (e.g., prep notes for every meeting; continuous vendor price audits).
Horizontal distribution plus role-aware recommendations can beat choosing one vertical too early.
Instead of verticalizing upfront, Town gathers patterns from power users by role/industry, builds reusable routine templates, and uses onboarding signals to recommend the right automations—moving toward highly personalized “80/20 custom software.”
WORDS WORTH SAVING
5 quotesBut post-AI, it just, it just, this was not gonna be a venture scale company.
— Jean-Denis Greze
There’s a ticking clock on the wall, which is how quickly can you get real word of mouth network effects and market penetration before the incumbents who have other advantages.
— Jean-Denis Greze
Humans are slow routers.
— Alex Rampell
I think of our product as a token-allocation machine.
— Jean-Denis Greze
You can’t teach someone. You have to just show the value prop to someone and do it for them.
— Jean-Denis Greze
High quality AI-generated summary created from speaker-labeled transcript.
