Hiring your first AI engineer as a non-technical founder.

What should you decide before you start hiring?

Write down the problem, not the job title. "We want to use AI" is not a brief. "Support tickets take two days to answer and we want a first draft ready in seconds" is. A specific problem tells candidates what they would own, tells you which skills matter, and gives you something concrete to talk about in every interview. Also decide what you can offer that a large company cannot: ownership, scope, speed and a direct line to the founder. Those are what attract strong early engineers.

Which AI engineer do you actually need first?

Most startups building products on top of foundation models need an applied AI or LLM engineer: someone who builds retrieval, agents and evaluation around existing models and ships them to users. If you are pre-product and need one person to build the whole thing, you need a founding engineer with strong AI experience. A machine learning engineer makes sense when your product depends on training models on your own data. A research scientist is rarely the right first hire unless the research is the product.

How do you assess skills you can't judge yourself?

You do not need to understand every technical answer to run a good process. You need structure and borrowed judgment. Ask a trusted technical advisor, an engineer from your investor's network or an experienced friend to run or join the technical interview. Use the same questions and the same scoring rubric for every candidate, and write down evidence before you compare notes. Our list of AI engineer interview questions that test judgment is a good place to start. Ask candidates to walk you through something they shipped: what problem it solved, how they knew it worked, and what they would change. Good engineers can explain their work to a non-technical listener. If a candidate cannot, that matters, because you will be working with them every day.

What are the red flags in an early AI hire?

• They talk about models and tools but never about users or outcomes.
• They cannot describe how they measured whether something worked.
• Every past project was a prototype or a demo, with nothing in production.
• They need a large team or a lot of structure to be productive.
• They make confident promises about timelines for problems nobody has scoped.

What should you prepare for the first 90 days?

Agree on outcomes before the offer, not after. A useful shape: in the first 30 days the engineer understands the problem, the data and the users, and ships something small. By 60 days a first version is in front of real users with a way to measure quality. By 90 days they have a view on what to build next and what the team needs. Give them access to data, users and decision-makers from day one. Most early AI hires stall on access, not ability.

Should a non-technical founder use a recruiter?

If you have a strong technical network, you may not need one. If you do not, a recruiter who understands AI roles can save you from the most expensive mistake: hiring someone who interviews well but cannot ship. At PeopleNotResumes we use AI to source widely, assess candidates on behaviour and decisions rather than resume keywords, and a human approves every candidate decision. For the wider process, read our guide on how to hire AI engineers.