The Wrong Person to Lead AI

A small business decides it needs an AI lead. The role has not existed before. The founder looks around the room and picks a person, almost always within five seconds of having the thought.

The person picked is one of two types. Either the most technical person in the building — the engineer, the analyst, the data-fluent operator — or the most junior person, on the theory that they're the digital native and will figure it out. Both choices are made with the same confidence. Both are usually wrong.

I want to be careful here, because I don't think either choice comes from a bad instinct. They come from a misunderstanding of what AI work actually is.

The technical person seems like the obvious pick. They know APIs. They aren't intimidated by JSON. They have already played with the tools. Surely they should lead.

The problem is that AI work inside an SMB is not, primarily, a technical problem. The technical layer — picking a model, writing the prompt, wiring the API — is the smallest and easiest part. The hard parts are: which workflow to automate first, what good looks like in this business, how to roll a tool out to a team that didn't ask for it, how to talk to a customer about a change to how their account is handled. None of these are technical questions.

When you put a technical person in charge of AI, they default to the part of the job they know how to do. They optimise the prompt. They benchmark the model. They build the pipeline. They do not, typically, sit with the head of sales for three hours to understand how she actually closes deals — because that is not the work they were trained to do. The result is an AI deployment that is technically excellent and commercially irrelevant.

The most-junior pick comes from a different but equally flawed instinct: AI is new, the juniors know new things, therefore the junior should lead. This was already wrong with social media. It is more wrong with AI.

Leading AI work in a business is not a question of comfort with the tools. It is a question of knowing what the business should be trying to do with them. It requires understanding which workflows matter, which customers matter, which mistakes are recoverable and which are not. The junior person, however bright, has not been in the room for the conversations where those judgments were forged. They do not yet know what good looks like for this specific business. Asking them to lead is asking them to direct a film without having read the script.

They will also struggle to influence the people they need to influence. Senior people on the team will not deeply change how they work because someone two years out of university told them to. That is unfair. It is also true.

The right person to lead AI inside an SMB has, in my experience, four traits.

They have run an operational function in the business for at least three years. Sales, ops, finance, customer success — somewhere they had to deliver outcomes through other people. They know how the business actually runs, not how the org chart says it runs.

They have credibility with the senior team. Other senior people listen to them. They can walk into the CFO's office and ask why a process exists, and the CFO will tell them rather than brush them off. This is not a technical credential. It is political and social.

They are curious enough to be useful, technical enough not to be intimidated. They do not need to write the prompt themselves, but they need to know what a prompt is, what a model is, what it can and can't do, and what the limits look like in practice. Most non-technical operators get there in a week if they apply themselves.

They are willing to be wrong in public. AI work involves trying things that don't work, in front of people whose respect they care about. The right person can do this without it eroding their authority — because their authority comes from operational track record, not from never being wrong.

If you pick correctly, you will notice the time allocation shifts. The right person spends maybe twenty per cent of their time on tools, models, and prompts. The other eighty per cent is spent in conversations — with the team about how they currently work, with customers about what they actually value, with the founder about what good looks like and where the line is.

That ratio is the tell. If your AI lead is spending most of their time on the technology, you have picked the wrong person, regardless of how good the technology gets. If your AI lead is spending most of their time on conversations that turn into clearer briefs and better-targeted work, you have picked the right person, regardless of how technical they are.

The good news is that the right person is usually already in the building. The bad news is that the obvious pick — the technical one or the junior one — is almost never them. If you've already made the obvious pick, it is not too late to change. The cost of changing now is much lower than the cost of running a misaligned AI function for a year.