Your AI Lead May Not Be Your Most Technical Person

“Prompting can be taught. Trust and rigour cannot be bolted on.”

I sat down recently with the leadership of a large New Zealand business. Careful operators. Not given to chasing shiny things.

They had made the sensible first moves on AI. A basic policy. A simple strategy. Some leadership oversight. So I asked the obvious question. Who is actually leading this?

The answer was surprisingly small. One person, supported by a colleague from an audit and assurance background.

They had technical capability available. But technical depth was not the main reason for the choice.

They wanted someone known for rigour, credibility and a willingness to test assumptions thoroughly.

Many people might see this as a category error. AI is a technology, surely it belongs with the technologists. But the more they explained it, the more it looked like the smartest move in the room.

Early-stage AI in a cautious business is not primarily a technical problem. It is an assurance and judgement problem. The hard questions are not “can we build this.” The hard questions are the uncomfortable ones. Can we trust this output? Where is the risk hiding? Who is accountable when it gets something wrong and no one notices for a month? Those are audit questions, and a good assurance mind asks them by reflex, on everything, whether or not anyone asked them to.

I have watched what happens when nobody does.

I have seen a business put a machine-learning tool into a workflow where staff were personally measured on the numbers it produced. The staff rejected it. Not because it was always wrong, but because they were being asked to stake their own accountability on a number they could not explain, challenge or independently verify. When the tool got something visibly wrong, that was the end of it. Trust, once gone, does not come back with a better accuracy score.

That pattern shows up across businesses. Finance, operations, hiring, client work. The moment you ask a person to put their name to an output they do not understand and cannot interrogate, they will stop using it. Correctly. In a cautious organisation, the barrier is rarely whether the technology works.

It is whether the people expected to rely on it trust how it was tested.

Which is why the person who leads it matters more than the tools they choose.

Their job is not to build the system. It is to build the conditions under which the organisation can justifiably trust it. In practice that means a few things:

• Ask where it could fail, before it does.

• Make ownership and checking explicit: who owns each output, who checks it, what happens when it is wrong.

• Make sure the people relying on it can actually challenge and verify it, not just take it on faith.

• Require evidence before confidence, not a slick demo.

Notice what that is not. It is not making people believe what the machine says. It is the opposite. The owner makes the organisation willing to engage with AI precisely because people trust that someone rigorous has pressure-tested it, mapped where it could fail, and will say so plainly when it cannot be relied on. People are not being asked to trust the machine. They are being asked to trust the process standing behind it.

And this is not a handbrake. Their job is not to slow the organisation down. It is to help good AI ideas move with enough evidence, ownership and transparency that people will actually use them. The assurance-minded owner is what lets a careful business move faster, because the organisation stops second-guessing every output and starts trusting how it was checked.

None of this means excluding your technical people. Some of them are also trusted, rigorous and commercially sharp, and if so they are strong candidates. The point is the criteria, not the job title. Choose the lead because they are trusted and rigorous, not simply because they are the most technical person available. The technical work can usually be supported. The judgement cannot.

There is also a practical catch that many organisations underestimate. Nobody wants to add headcount, and the person you hand this to still has a day job. It is tempting to call the role a priority and leave it there. But if that person keeps a full workload, leadership has not assigned ownership. It has assigned an extra responsibility without the capacity to carry it, which produces part-time governance over full-time risk. Real ownership means funding the capacity to do it properly.

So the question to sit with is not “who is our most tech-savvy person?” The person you want may not be highly experienced with AI yet, and that need not disqualify them. They can learn the tools, and specialists can support the build. But the credibility to challenge a senior leader, to surface bad news early, and to earn the organisation’s confidence usually takes years to build. That is the part you cannot hire in for a project.

Do not choose the person most excited about what AI can build. Choose the person your organisation trusts to say when it should not be built at all.

Find out where your AI ownership risk really sits

Most businesses have people using AI, but no clear owner for the risk, decisions or outputs.

That creates gaps in oversight, accountability and trust.

Take the 5-Minute AI Readiness Snapshot

--

About the Author 

Mark Lucas is Co-founder at Generation AI, where we help New Zealand business leaders build real AI capability.

Information provided is general in nature and does not constitute legal advice