If you are looking for a CTO for an AI company, the questions that predict fit are not about credentials. They are about what the person has actually shipped to production, whether they have real AI-specific skills like building evaluations and data pipelines, how they think about equity and vesting, and whether they can carry ambiguity. A wrong technical co-founder usually costs you around a third of the company and is painful to undo, so vet before you commit.
A co-founder search is the most consequential hire you will never call a hire. You are not filling a role, you are choosing a partner who will own the hardest part of the company and a large piece of the equity, often something like a third. Most founders vet for the wrong things: a big-company logo, a prestigious degree, a strong interview. Those predict very little. The mechanics of how founder equity and vesting are typically structured are worth understanding before any conversation, but the harder part is judging the person. Here is the checklist that actually matters when you are looking for a CTO for a vertical AI company, and the alternative worth weighing before you give away that third.
Shipping history: what have they actually put in production
Start here, because it predicts more than anything else on a resume. Ask what they have personally built and shipped to real users, at what scale, and what broke. You want specifics: the system they owned, the load it carried, the incident they got paged for at 2 a.m. and how they fixed it. A candidate who can walk you through a production system they were accountable for, including its failures, is telling you they can own yours. A candidate who describes architectures in the abstract, or whose experience is research or prototypes that never carried real traffic, may be brilliant and still not the person to run your product. Build-versus-research is the single most common mismatch in AI hiring: the skills that win a benchmark are not the skills that keep a product up.
AI-specific skills: the ones generic seniority misses
A strong general engineer is not automatically equipped to build a vertical AI product. The discipline has its own toolkit, and you should probe for it directly rather than assume seniority covers it.
| Skill area | The real question | What a weak answer looks like |
|---|---|---|
| Evaluations | How do you measure whether the AI output is good enough to ship? | No systematic evals, ships on vibes |
| Data | How do you source, clean, and use proprietary data as an advantage? | Treats data as an afterthought |
| Model strategy | When do you fine-tune, when do you prompt, when do you use a smaller model? | Defaults to the biggest model for everything |
| Agents and reliability | How do you make an agentic workflow reliable and debuggable? | Has only built demos, not durable systems |
| Cost and latency | How do you keep inference cost and latency viable at scale? | Never modeled unit economics of inference |
You do not need to be technical enough to grade the answers perfectly. You need to hear whether the person reasons about these tradeoffs at all. Someone who lights up on evaluations, data advantage, and inference cost is thinking like an operator of an AI product. Someone who waves them away is likely to learn expensive lessons on your equity.
How they think about equity and vesting
The equity conversation is a character test disguised as a negotiation. A serious technical co-founder expects to earn a large stake and also expects that stake to vest over time, typically four years with a one-year cliff, so that if the partnership fails early, the equity does not walk out the door. A candidate who wants a large slice with no vesting or resists a cliff is showing you how they will behave when things get hard. That is not a founder protecting their upside; it is a partner refusing the mutual accountability that vesting exists to create. How a person approaches this, whether they think in terms of shared risk or personal protection, tells you more about the next five years than any technical answer.
Values, decisions, and conflict
You will disagree with this person under pressure, so learn how they handle it before you are married to them. Ask about a real conflict with a past co-founder or manager and listen for whether they take any responsibility. Ask how they make a decision when the data is ambiguous, because most early decisions are. Work together on something real before committing, a paid trial project or a few weeks building side by side, because a working relationship reveals what interviews hide. The co-founder split is one of the hardest relationships in a company to undo, so the cost of skipping this diligence is enormous.
The alternative: a team instead of a single bet
Here is the reframe most founders miss. Even a perfect vetting process is a bet on one person carrying the entire technical side of the company, and one person is a single point of failure for the thing you cannot build yourself. There is another path. Instead of finding, vetting, and marrying one CTO, you can co-found with a venture builder that provides an institutional technical cofounder and a full founding engineering team from day zero. The five real alternatives to the co-founder search, and what each costs, are laid out in technical co-founder alternatives for an AI startup.
gAI Ventures co-founds vertical AI companies in financial services, enterprise productivity, and commerce, and supplies the engineering team led by its own CTO rather than leaving you to find one. The equity math is different too: many arrangements hand a single co-founder around a third, while gAI keeps the fund and operating company combined near 20 percent for a cleaner cap table, the tradeoff examined in venture studio economics. It validates the idea in a four-week sprint before anyone commits, which is its own form of diligence, described across the gAI Ventures blog and in the gAI Ventures manifesto. This is not the right answer for everyone. If you have found a genuinely great technical co-founder and can survive the search, that partnership has the highest upside there is. But if you are still looking, weigh the team against the single hire before you give away a third.
How gAI decides what to build
The reason the team-first model works is that the hard part of a vertical AI company is not only engineering, it is picking the right problem with the right operator, which is why the sectors and theses gAI co-founds against are published as our vertical AI investment theses, the companies already built are on the gAI Ventures portfolio, and the operators are on the gAI Ventures team. A founding team assembled around a validated idea starts from a different place than a single co-founder hired on conviction alone.
Frequently asked questions
- What should I look for in a CTO for an AI startup?
- Prioritize what they have shipped to production over their pedigree. Ask for specific systems they personally owned, the scale those ran at, and how they handled failures, because a production track record predicts fit far better than a degree or a famous employer. Then probe for AI-specific skills that generic seniority can miss: building evaluations, using proprietary data as an advantage, deciding when to fine-tune versus prompt, making agentic systems reliable, and keeping inference cost and latency viable. Finally, watch how they think about equity, vesting, and conflict, since those predict the partnership more than any technical answer.
- How much equity does a technical co-founder or CTO get?
- A technical co-founder who joins early and builds the product typically earns a large stake, often in the range of a third of the company, reflecting that they are a full partner taking real risk. A CTO hired later, after the company and some funding exist, usually receives a smaller equity grant plus salary. Whatever the size, the equity should vest, commonly over four years with a one-year cliff, so that an early split does not permanently give away ownership if the partnership does not work out. Resistance to standard vesting is a warning sign worth taking seriously.
- Should a technical co-founder's equity vest?
- Yes. Standard practice is vesting over about four years with a one-year cliff, meaning no equity is earned in the first year and the rest accrues monthly after that. Vesting protects the company and both founders: if someone leaves early or the partnership fails, unvested equity returns to the company rather than walking away with a departed founder. A candidate who insists on a large unvested stake is asking you to take on risk they will not share, which tells you how they are likely to behave when the company hits its first hard stretch.
- Is it better to hire one CTO or use a venture builder's team?
- It depends on whether you have already found the right person. A single great technical co-founder has the highest upside, but you are concentrating all technical risk in one relationship that is hard to undo, and the search can take many months. A venture builder provides an institutional technical cofounder and a full founding engineering team from day zero, spreading that risk and often keeping the cap table cleaner, but you co-found within its model and sectors. If you have a proven co-founder in hand, take it; if you are still searching, weighing the team against the single hire is worth doing before you commit a third of the company.
- How do I vet a technical co-founder before committing?
- Work together on something real first. A paid trial project or a few weeks of building side by side reveals how someone communicates, makes decisions under ambiguity, and handles disagreement, which interviews cannot. Ask directly about a past conflict with a co-founder or manager and listen for whether they take responsibility. Check references from people who have shipped alongside them, not just managers. And confirm alignment on the practical terms, equity, vesting, roles, and decision rights, in writing before you commit, because ambiguity there becomes the fight that breaks companies later.
End of article · #003