Diligence on an AI Startup

Joining a small AI company is a concentrated bet, and the diligence is a normal part of a senior conversation. Founders expect it. Being uncomfortable asking is common and expensive.

Money. How is revenue trending, what's the runway in months, when do you expect to raise again and what needs to be true for that round. You won't always get numbers, but the shape of the answer — precise, vague, or evasive — is itself information. "We have eighteen months and we're default-alive at current burn" and "we're in a great position" are very different answers.

The bet. What's defensible when the model providers ship something adjacent to your product? A lot of AI companies are one release away from being a feature. Good founders have thought about this hard and will tell you — proprietary data, workflow depth, distribution, an integration surface. A founder who hasn't considered it is the signal, not the answer.

Cost structure. Gross margin on an AI product is a live question: inference cost per customer versus what they pay. "What does a customer cost us to serve?" is a question few candidates ask and every good founder can answer.

Customers. How many, how concentrated, are they paying or piloting. Ten paying customers beats a thousand free ones, and "we're in a pilot with a large enterprise" often means nothing yet.

Equity, in detail. Percentage of fully diluted shares, not just the option count. Strike price and current 409A or equivalent. Vesting and cliff. Exercise window after leaving — 90 days is common and can make the grant worthless. Preference stack, because a large liquidation preference means common shares can be worth nothing in a decent exit. Get the answers in writing.

Cross-border. Employer of record versus contractor, and what that does to equity treatment and your tax position — see remote across US and Europe.

Equity in an early-stage company is most of the theoretical upside and usually worth zero; the salary and the learning are what you can count on. Knowing which bet you're taking, and pricing it deliberately, is the difference between a considered decision and hope. Asking these questions also signals that you've been near a startup before.

• Ask directly about runway in months and what the next raise requires; evasiveness is the answer.
• Ask what's defensible if a model provider ships something adjacent — the absence of a considered answer is the red flag.
• Ask what it costs to serve a customer; inference cost decides whether the business works.
• Distinguish paying customers from pilots and free users.
• Equity means percentage fully diluted, strike, 409A, vesting, cliff, and post-departure exercise window.
• A 90-day exercise window can make a grant unusable — extended windows exist and are worth asking for.
• Ask about the liquidation preference stack; common shares can be wiped out in a mid-size exit.
• Price equity at approximately zero and decide whether the salary, the work, and the learning are worth it on their own.