What Happens When Al Agents Can Spend Your Money?

What It Covers

a16z's Alex Rampell interviews Jean Denis Greyes, former Plaid CTO and Town founder, about building AI assistants for workplace workflows. They cover Town's pivot from SMB tax software, why model neutrality matters for enterprise buyers, how agents will reshape commerce and software distribution, and why AI adoption remains concentrated among technical tinkerers rather than spreading broadly across businesses.

Key Questions Answered

• •Pivot timing in cyclical businesses: When building in annual-cycle markets like tax software, founders have a narrow window to change direction — roughly one month after each season closes. Miss that window and you're locked in for another year, burning capital without the option to redirect. Jean Denis identified this constraint and used it as forcing function to pivot Town away from tax prep before the next season began.
• •Model neutrality as enterprise sales strategy: When pitching mid-market firms of 500–1,000 employees, Town positions itself as model-agnostic — routing to whichever frontier model performs best at any given moment. This directly addresses a real enterprise fear: companies that committed exclusively to OpenAI two years ago found themselves at a disadvantage when Anthropic led on coding tasks for roughly 12 months, forcing costly dual-vendor arrangements.
• •PLG before enterprise to protect product quality: Starting with product-led growth for SMBs and mid-market forces teams to optimize for the actual user, not the procurement buyer. Town's co-founder, a Dropbox veteran, observed that world-class consumer products can be adapted upward into enterprise, but enterprise-first products almost never achieve consumer-grade experience. Compliance-first design creates a ceiling on product quality that compounds over time.
• •Tinkerer gap explains uneven AI adoption: AI capabilities sufficient to automate meaningful business tasks have existed for 12+ months in tools like Claude and ChatGPT, yet adoption remains concentrated near engineering-heavy environments. The bottleneck is not intelligence — it is the absence of tinkerers who can identify and configure automations. Town's core product challenge is delivering tinkerer-equivalent value to non-technical users without requiring them to prompt-engineer their own workflows.
• •Token allocation as product framework: Town frames its product as a token allocation machine — surfacing opportunities where spending compute delivers measurable ROI, then letting users decide. A concrete example: running a monthly price-comparison across wholesaler PDFs costs roughly one dollar per month but can surface hundreds or thousands in savings during the one month per year when prices shift significantly enough to warrant switching suppliers.
• •Data ownership as switching-cost inversion: Town sells enterprise customers on owning their full context layer — all memories, preferences, and business knowledge accumulated through AI interactions — so that switching to a competitor would take one day. This inverts the typical SaaS lock-in model. Jean Denis frames this as a deliberate competitive stance: if Town cannot maintain the best product, customers should be free to leave, and that credible commitment itself becomes a differentiator against lab-owned vertical products.