What a CHRO Should Do First with AI (Without Handing It to IT)

Hook

Here is the quiet fear, said plainly. AI lands on the executive agenda, and the CHRO who is not visibly leading it watches it get handed to IT, to Legal or to a freshly minted Chief AI Officer. The work of the function continues. The ownership of its future moves down the hall. The seat does not get cut in one meeting. It shrinks, one decision at a time, until people-strategy is something done to HR rather than by it.

If that is the worry, the good news is that the fix is entirely in your hands, and the first move is smaller than you think. It is not a transformation program. It is not a vendor selection. It is not a new headcount. It is a single week of deliberate, hands-on work that puts you, not a proxy, at the center of how your organization uses AI on its people.

This lesson is that first move, laid out as a plan you can start now. The goal is not to make you technical. It is to make you the person in the building who clearly owns the human-AI agenda, because you have done the work yourself, on real HR problems, before anyone else in the C-suite got there.

Context

This anxiety is not niche and it is not in your head. In a 2026 survey of about 150 CHROs, 91% placed AI and the digitization of the workplace in their top five concerns. Almost every CHRO is feeling exactly what you are feeling. That is not a reason to panic. It is a reason to move, because in a crowded field of anxious peers the few who act first will be visible.

The instinct under pressure is to delegate the hard, technical-sounding thing to whoever owns technical things. It feels responsible. It is the trap. The moment AI on your people, your hiring, your performance data, your workforce planning, is owned by IT or Legal, two things happen. You lose the agenda, and you also lose the part that was always yours: judgment about people. IT can run a model. Legal can flag a risk. Neither can decide what good looks like for your workforce. That is the CHRO's job and it does not transfer, unless you give it away by staying on the sidelines.

The same survey found that 47% of CHROs have not yet established clear productivity measurements for the AI they are already using. Read that together with the first stat and the picture is sharp: nearly everyone is worried, many are already deploying, and roughly half cannot say whether it is working. That gap between activity and evidence is precisely the space a CHRO is built to lead, if you step into it now.

So the question is not whether to engage. It is how to engage in a way that puts you, credibly, at the front. The rest of this lesson is the answer, in five steps you can start this week. None of them requires you to become technical. All of them require you to stop delegating and start doing.

That is the posture. Not the CHRO who commissioned a study. The CHRO who used the tool visibly, on real work, and pulled her team forward because they saw what became possible. That is what owning it looks like, and it starts with your own hands on the work.

Steps

Step 1: Use AI yourself on a real HR problem, this week

Before you lead anyone, do the work once. Pick a genuine, tedious-but-important piece of your own week: a workforce-planning memo, a messy engagement-survey readout you need to turn into a board narrative, a first-draft policy, the strongest case for and against a reorg. Bring it to Claude or ChatGPT and brief it like a sharp new analyst: who you want it to be, the context, the task and the format you want back.

If you have never done a focused first session, do that one first. The companion lesson "Your First Hour with Claude, for Busy Executives" walks the exact pattern, step by step, and it is the fastest way to earn the credibility this whole plan rests on. You cannot lead a capability you have never personally used. One real session changes that.

Step 2: Name the people-side outcomes only you can own

Now decide what good looks like, in your domain, in business language. This is the part no other function can do for you, and the part that anchors your claim to the agenda.

Write down three to five outcomes where AI touches people and the judgment is inherently yours. For a CHRO, the strong candidates are concrete:

• Faster, fairer hiring without amplifying bias in screening
• Manager capability built at scale, not just for the top team
• Workforce planning that uses better signal, not just more dashboards
• Employee questions answered accurately and privately, at any hour
• Performance and development conversations made better, never automated away

For each, state the outcome and the line you will not cross. "We will use AI to speed up first-pass resume review, and a human makes every advance-or-reject decision." That sentence is governance and it is strategy, in plain English, owned by you. IT cannot write it. Legal can only check it. You define it.

Step 3: Choose and run one scoped pilot, on your terms

Pick exactly one of those outcomes and run a single, contained pilot against it. Not a program. One scoped use of AI on one real, recurring HR pain, with a measure of success defined before it starts and a fixed window to judge it.

The discipline here matters more than the scale. A good first pilot is visible enough that success is noticed, contained enough that success is achievable, and tied to a real recurring pain so the result means something. Write the measure down first: hours saved per manager, time-to-first-interview, the share of employee questions resolved without escalation. Defined beforehand, that number is evidence. Invented afterward, it is a story. (The track's lessons on the adoption roadmap and the business case go deep on choosing and measuring pilots; this is the CHRO-owned starting point.)

Critically, you sponsor it. Not IT running a pilot you observe. You choosing the problem, setting the measure and owning the result. That is the difference between leading AI and being briefed on it.

Step 4: Set the guardrails before you scale, in plain language

This is where most CHROs are tempted to hand the whole thing to Legal, and where you must not. Legal checks the rules. You decide the principles. Write three or four guardrails a non-lawyer can understand and a busy manager can actually follow:

• What employee and candidate data may and may not go into a tool, and which tools are approved
• Where a human decision is mandatory and cannot be delegated to a model
• What gets disclosed to employees and candidates about how AI is used on them
• Who people ask when they are unsure, and that asking is encouraged, not punished

Plain-language guardrails are not a substitute for Legal review. They are the thing Legal reviews. You write the people-first version, Legal pressure-tests the legal exposure, and the result is governance that is both compliant and humane. That collaboration, with you holding the pen, is the model. It is the opposite of handing it over.

Step 5: Make your ownership visible to the C-suite

The final step is the one that protects the seat. Having done the work, claim the ground out loud. In the next executive meeting, do not present a request for a study. Present a position: here is how I am using AI on our people, here is the pilot I am running and its measure, here are the guardrails I have set and here is the agenda I own for the next two quarters.

This is not politics. It is clarity. When AI on people has a clear owner who has visibly done the work, the CEO stops looking for someone else to hand it to, and the new Chief AI Officer (if one arrives) becomes your partner on infrastructure rather than your replacement on strategy. The seat does not shrink when you are obviously the person leading the part that was always yours.

Then do what Kim Pecina did: teach your own team, in the open, because they have seen what you can do with it. Demonstrated capability pulls a function forward in a way no mandate ever does. That is how a CHRO closes the gap inside their own house and ends the year more central to the business, not less.

Recap

• The fear is legitimate and near-universal: 91% of CHROs rank AI in their top five concerns. The seat shrinks not in one meeting but one delegated decision at a time.
• Handing AI to IT, Legal or a new Chief AI Officer keeps your title and loses your agenda. Owning it does not mean coding. It means doing the people-judgment work yourself, first.
• Step 1: use AI on a real HR problem this week. You cannot lead a capability you have never used. Earn the credibility the rest of the plan rests on.
• Step 2: name three to five people-side outcomes only you can own, each with the line you will not cross. That is strategy and governance in plain English.
• Step 3: run one scoped pilot you sponsor, with the measure of success written down before it starts, not after.
• Step 4: set plain-language guardrails yourself, then have Legal pressure-test them. You hold the pen.
• Step 5: make your ownership visible to the C-suite and teach your own team in the open. A clear owner who has done the work does not get routed around.

You now have the first move that keeps the people-AI agenda yours. Where you go next depends on where you are.

If you have not yet made AI part of how your function runs, the operating-model lesson turns these first steps into a standing rhythm: ownership, policy, a sanctioned toolset and a sharing loop.

If you are already deploying AI across a real team and your problem is scaling it safely, not personal credibility, go straight to the deployment protocol: build-versus-buy, the pilot-to-production gap, governance and measuring ROI across the function.

Continue: The AI Operating Model →

For already-deploying operators: The Enterprise AI Deployment Protocol →