Hiring the first AI team in a non-tech company

The usual mistake

A company decides it needs AI capability and posts a dozen roles at once. Months later it has a team with strong credentials, no clear problem and little connection to the people who run the business. The better path is slower at the start and faster after: begin with one problem, one owner and one small team that sits next to the work.

Who to hire first

The first hire should be a senior builder who has taken a system from idea to production in a real organization. Look for someone who is comfortable talking to a plant manager or a finance lead, who understands that the model is a small part of the system, and who can explain trade-offs without jargon. Titles vary. What matters is judgment about scope, data, reliability and people.

The second and third hires depend on the first problem. Common needs are an engineer who builds and runs the production system, a data specialist who owns quality and access, and an operator from the business who knows what a good outcome looks like and will adopt the result. That last role is often overlooked, and it is often the most valuable.

Where to put the team

Place the team close to a business owner who has a real target. Avoid a free-floating lab that reports to no one with operating responsibility. At the same time, avoid scattering individual hires across departments, where each is isolated. A small central team that works in embedded stints with business units is a practical middle path for many companies.

What to measure in the first year

• Number of workflows moved to production and still in use.
• Time from idea to first working version.
• Share of work in the target workflow completed without manual rework.
• Whether business owners ask for the team's help without being told to.

Notice what is missing: model accuracy on benchmarks and headcount growth. Those matter later, if at all.

Hiring checklist

• Can they describe a system they shipped and what went wrong?
• Do they ask about the data and the users before they ask about the model?
• Can they say no to a project with a clear reason?
• Will they write down what they build so others can maintain it?
• Do they explain concepts in plain language to a non-specialist?
• Would operators in your company want to work with them?

Build, borrow or both?

Few non-tech companies can recruit a full team quickly, and the market for senior people is competitive. A blended approach is common: bring in outside specialists to start, then hire to own the result. If you use outside help, insist on documentation, shared code ownership and a plan for handover. The decision framework in build, buy or partner for models applies to people as much as to technology.

Retention and growth

People who build AI systems in an unfamiliar industry stay when the work is meaningful, the data is accessible and the leadership listens. They leave when projects are cancelled without explanation, when they are asked to justify every decision to a committee, or when their work is hidden behind a vendor. Give the team direct access to operators, a protected budget for experiments, and a clear path for what happens after a pilot succeeds.

A hypothetical first year

Interview questions that reveal judgment

Ask candidates to describe a project they stopped, and why. Ask how they would test a system when no labeled data exists. Ask what they would do if an executive wanted a feature that the data could not support. Ask them to explain a technical trade-off to someone who does not work in technology. Strong candidates answer concretely and admit uncertainty. Weak answers lean on tool names and impressive vocabulary.

Pay, titles and expectations

Expect to pay market rates for experienced people and to explain a career path they cannot see in a conventional company. Be clear about decision rights: what the team may build alone, what needs approval, and how priorities are set. Avoid titles that promise more authority than the company is prepared to give.

What to ask

If you are the owner of an established company, ask yourself before hiring: which workflow will this person own in the first ninety days? Who in the business will they work beside? What would make us stop? If those answers are unclear, a hire may be premature, and a conversation may be more useful. Our page for companies describes how we think about this alongside owners and operating teams.

This article is for general information only and is not investment, legal, tax or accounting advice.