Somewhere in the last year the ask changed. It used to be “keep us safe.” Now it is “keep us safe, and also, why aren’t we doing more with AI?” Same person, both jobs, and the second one usually arrives in a hallway conversation with no budget attached.

Here’s the thing. That is not a contradiction. It only feels like one because the AI question keeps getting asked in the wrong shape. “Can we use AI?” is a yes or no question, and yes or no is where a security leader gets stuck. Say no and you are the department of no. Say yes and you own whatever happens next.

So change the shape of the question. The version you can actually answer is: what can AI reach, what is it allowed to do, and how would we know? Those three are answerable today with what you already have, and the answers turn a vague mandate into a plan.

Think of it like a new hire on day one. You don’t hand them every key in the building and hope for the best. You give them a badge that opens specific doors, you tell them what they can sign for, and you keep a log. Nobody calls that being anti-hire. That is onboarding. AI needs onboarding.

A few things that have worked for the security leaders I talk with:

Write the one page before anyone asks for it. Three sections: what is fine today (which tools, which data), what is on the path with conditions, and what is off limits for now and why. Keep it in plain English. If Karen in accounting can’t read it in two minutes, it will not change anyone’s behavior.

Say yes to one thing out loud. Pick a low-risk use case, something like drafting internal documents from non-sensitive material, and approve it publicly. That single yes buys more credibility than a dozen careful maybes. It also gives people a sanctioned path that can help reduce shadow AI.

Move to the front of the line. Ask to be part of intake, not just approval. When a team is scoping an AI project, the cheapest moment to set boundaries is before the vendor demo, not after the pilot has real data in it.

Bring a number, not a feeling. Can your team say today what data AI can already reach? For most of the security leaders I talk with, the honest answer is “not yet,” and that is not a failing. It is a measurement nobody asked for until now. Go find yours. A measured gap is something a board can fund. A worry is not.

Borrow the skills before you hire them. The mandate landed on a team that was already short-handed, and nobody added headcount with it. For the first stretch, the inventory, the one page, the first pilot, borrow the specialized skills you are missing for a defined window instead of waiting on a requisition. Scope that person the way you would scope the AI itself: named systems, least access, an end date. Make knowledge transfer a deliverable, so what they learn stays with your team when they leave.

Decide who approves the actions that matter. Reading is one thing. Writing to a system, sending an email, moving money, changing access: those get a human sign-off, and you decide which ones now, while the list is short.

None of this is exotic. So basically it is the same discipline you already apply to a new SaaS vendor or a contractor with a laptop, pointed at a technology that moves faster and talks back.

The key thing is that you get to set the terms, but only if you set them first. The pressure to do AI is not going away. What you can change is whether the conversation starts with your one page or with someone else’s screenshot of a demo.

Practical next step: this week, write the three-section page. A rough one is fine. Then ask one question of every AI proposal that lands on your desk: what can it reach? If nobody can answer that, you have found your first project.

Where Ovatio fits

The AI Readiness Assessment helps establish what is already in use, where boundaries need attention, and which decisions to make first within an agreed scope. If the team is short-handed for the first stretch, contract IT staffing covers a defined window with a defined scope.