Jean-Denis Greze kills Town's tax-prep product after it neared $1M in first-year revenue, blocked by the annual SMB sales cycle.
Town's email and calendar assistant now adds over $1M ARR monthly and has tens of thousands of users.
Greze says product advantage now lasts months, so Town differentiates through proactivity, lab neutrality, and marketing to Midwest recruiters.
Town killed a $1M tax AI to build an assistant
Town's founder killed a tax-prep AI business approaching $1M in revenue because annual sales cycles for small businesses made hyper-growth unrealistic. He rebuilt around email and calendar, shipping a proactive assistant that now adds over $1M of ARR a month. Greze thinks assistant capabilities will converge, so product advantage lasts months and user choice, proactivity, go-to-market and brand matter.
Takeaways
Product advantage now lasts only months. Capability differences will disappear as harnesses normalise, so which users you pick, proactivity, go-to-market and brand matter more. Greze aims to be the only funded startup, and Town verticalises marketing for non-Silicon Valley recruiters.
Only 15% of people who try Town pay. About 85% of people who try the product do not like it, and 15% pay. Growth is still mostly word-of-mouth, though Town has started paid marketing. Users who built their own version switched to Town because upkeeping it takes effort.
Town hires tinkerers who talk to customers. The interview starts with architecture and previous-work stages, then a two-and-a-half to three hour in-person build in a six-month-old snapshot of Town's code base using an agent. Greze judges candidates on when they read the code and believes in firing quickly.
Agent payments will run on cards, not stablecoins. Akhund is sceptical of stablecoins and prefers cards with limits because you can dispute them and change card numbers. Greze says card fraud signals are stronger than stablecoin or ACH, and Akhund says a bounty marketplace would build card payments in anyway.
Willingness to pay and NRR matter early. Greze watches willingness to pay and net revenue retention from its seven-day trial. Akhund says gross profit matters most, and retention early on. Town looks at a session only with the user's permission and does not run aggregate evals.
Vertical AI needs a legal reason to exist. Greze argues vertical AI will not succeed in this generation unless there is a legal or compliance reason for it to exist, because capabilities are the same. Town stays horizontal and lab-neutral for businesses of 500 to 2,000 employees.
How Town pivoted from tax prep to email
Tax AI reached close to $1M of revenue in year one, but the annual sales cycle for small businesses made $2-3M and hyper-growth unrealistic.
After the first tax season, Town told customers and partners it would not serve them the following year and helped the tax team find jobs elsewhere.
Tax preparers asked for meeting scheduling and help preparing for client meetings, which pointed to email and calendar.
Tap a name to go there · tap a line to see why they are linked
On this episode 4
Show 1
A three-day prototype shipped in under two weeks, drawing users through a sign-in button on a blank town.com page asking for Gmail access.
Town launched in June, went from about 15 to 30 daily active users over a holiday break, and has since added over $1M of ARR in each of the last two months.