A founder sees AI projects appearing across the business. Product is experimenting, operations has a few promising workflows, and customers are asking what the company will do next.
The obvious response can be to hire an AI lead. Sometimes that is exactly right. But the title is a poor substitute for a clear mandate.
Before you open the search, work out which problem the person would actually own. You may discover that you need a builder, a product leader, a data foundation or a better decision about where to start.
Find the constraint before naming the role
A team that cannot move a promising prototype into a reliable product has a different problem from a team that does not know which customer problem is worth solving. Both can ask for an ‘AI lead’ and mean entirely different work.
Write down one or two use cases, who they serve and what is preventing progress. Is the constraint data quality, product direction, engineering capacity, governance or ownership across teams? Be honest when it is more than one.
The answer should shape the role. A strong candidate cannot take responsibility for a result that the company has not defined.
Define an outcome you can inspect
‘Make us an AI company’ is not an outcome. A useful first mandate names a workflow or product result, a quality standard and the people who will judge it.
That might mean moving one customer-facing feature from trial to dependable use, or reducing a specific manual step without weakening review. The exact measure depends on the work. Include operating cost, exceptions and human checking, not simply output speed.
You do not need to pretend you know the complete roadmap. You do need enough clarity for a candidate to explain what they would test first and what evidence would change their mind.
Interview the decisions, not the vocabulary
Tool fluency is useful. Better evidence comes from a decision under real constraints: what the candidate built, what they chose not to automate, how they checked failures and how the work changed after deployment.
For a leadership hire, go further. How did they win support from the people whose work changed? Who carried responsibility when a result was wrong? What did the team learn and keep using without them?
Give candidates your real conditions. The state of your data, the limits on spend and the standard of review all belong in the conversation.
Know when to wait
If there is no agreed use case, no owner for the business result and no path to usable data, a senior hire may spend their first months negotiating a brief that founders could settle now.
A short discovery effort may be the better next step. Equally, if several teams are already making consequential AI decisions without shared ownership, waiting may create its own cost.
The right question is not whether you need an impressive AI title. It is whether this person will have a real problem, the authority to solve it and a way to show what changed.