Pre-seed AI investing is harder than it looks. The technology moves so fast that any thesis you have about a specific architecture or model capability is wrong within six months. The competitive landscape resets every quarter as model labs release new capabilities that vaporize entire startup categories overnight. So how do we underwrite a founder when the ground is moving under everyone's feet?
We've taken hundreds of pitches over the last twelve months. The pattern that separates the founders we back from the ones we pass has surprisingly little to do with technical depth. Three things matter more.
1. Workflow obsession
The best agentic AI founders are not the ones who can recite the latest model benchmarks. They are the ones who have spent years inside a specific industry, watching the same broken workflow play out thousands of times, and have decided they will not tolerate it for one more year. Their conviction is anchored in pain, not in technology. When the model layer changes, their thesis doesn't.
The clearest tell: ask a founder to describe a single customer's day, hour by hour. Workflow-obsessed founders give you the entire shape of the day in five minutes — who calls whom, where the spreadsheet lives, what breaks at 4pm on a Friday. Technology-first founders give you the model architecture and ask you what industry you'd like them to apply it to.
2. Iteration speed
Agentic systems are messy. The first version of any agent fails in surprising ways, and the only path forward is to ship, watch users, and ship again — sometimes twice a day. We back founders who treat product development as a continuous experiment, not a quarterly roadmap. The founders who build a beautiful prototype and then disappear for three months to perfect it almost always lose to scrappier teams shipping ugly versions weekly.
What we measure on iteration speed
- •Cycle time from customer feedback to shipped fix — best founders are under 48 hours.
- •Number of production deploys in the past 30 days — we want to see double digits, not zero.
- •Whether the founder watches their own product being used (recordings, logs, sit-alongs).
- •Willingness to kill a feature that didn't work, in public, the same week.
3. Commercial maturity
AI founders frequently underprice themselves and overinvest in product polish before they have a single paying customer. We look for founders who are charging from day one, even if they're charging $500 a month for a half-broken product. The willingness to ask a customer for money — and to listen carefully when they say no — is one of the most reliable predictors of long-term commercial success we've found.
Free pilots are a leading indicator that a founder is afraid of the answer. Real customers signal real demand by reaching for their card. If your prospect won't pay $500 to try, they won't pay $5,000 to keep using it. Find that out in week one, not month nine.
What we don't need
We also pay attention to what we don't need.
- •We don't need a finished product. A working demo and one paying pilot is enough.
- •We don't need an enterprise pipeline. We'd often rather see ten SMB customers than one logo.
- •We don't need a Stanford PhD or an OpenAI alumni badge. Pedigree is uncorrelated with our best returns.
- •We don't need a fully assembled team. A clear-eyed solo founder who knows their next two hires beats a five-person team with no conviction.
- •We don't need a complete deck. A loom video and a Notion page works.
What we need is a founder who knows their workflow cold, who has shipped something real (even if small), and who has at least one paying customer or a clear path to one within ninety days.
Our process
- First call within five business days. We read every cold email.
- Second call within seven days. We meet your co-founders, talk to a customer, and walk through the product live.
- Decision within fourteen days. Yes is yes; no is a real explanation.
- Wire within five business days of yes. No 'subject to syndicate' theatre.
If that's you, we move fast. First call to decision in two weeks. No process, no syndicates, no waiting for the next investment committee. We meet, we diligence, we wire. That's it.
