5 Things CHROs Must Do Before the Next AI Agent

TL;DR

Short answer: I think most CHROs are reacting to AI instead of shaping it — the ones who matter will have done five specific things before the next deployment, not after.

Key facts:

• Korn Ferry: 42% of CHROs are prioritizing AI investment, but only 5% of HR teams feel ready to implement it
• IBM's AskHR agent now handles 94% of HR queries (11.5M+ interactions in 2024); only 6% still need human judgment
• Grant Thornton: 78% of senior executives doubt they'd pass an AI governance audit within 90 days
• 22% of organizations have already stopped entry-level hiring; ~50% expect to freeze/cut it by 2027 — while IBM is tripling entry-level hires
• Forbes Tech Council: roughly half to two-thirds of employees use AI without company approval, senior leaders included

What this means for leaders: The five moves are: (1) audit which HR tasks AI will actually replace, (2) claim AI governance ownership before it defaults elsewhere, (3) redesign the entry-level pipeline instead of freezing it, (4) document the human-judgment skills every role needs post-automation, (5) set a real AI use policy before employees improvise their own.

There are two kinds of CHROs in this AI moment.

The first one hears that the CTO is deploying an AI agent to handle benefits questions and says "great, keep me posted." They show up to the all-hands after launch, answer the inevitable question about job security, and move on.

The second one was already in the room when the decision was made. They shaped the governance model, redesigned the impacted roles, and set the use policy before employees found a workaround.

The difference between them is not resources or org size. It is preparation. And right now, most CHROs are not prepared.

Why This Matters Now

Here's the gap that should concern you: Korn Ferry's global survey of 750 HR leaders found that 42% of CHROs are now prioritizing AI investment — but only 5% of HR teams feel fully prepared to actually implement it. That's not a rounding error. That's the distance between what's being asked of the CHRO seat and what most CHRO seats are built to deliver.

IBM's AskHR agent now contains 94% of all HR queries the company fields — more than 11.5 million interactions in 2024 alone. That means 94% of what Tier 1 HR used to do is now handled by an agent. The 6% that remained required human judgment. If your function looks anything like that model, you have 18 months before your headcount case changes permanently.

Anthropic's own labor market research draws the same line a different way: it distinguishes between what AI can theoretically automate and what companies have actually redesigned their workflows to use it for. That observed gap — capability sitting unused because nobody rebuilt the process around it — is not a safety net. It's a countdown.

The deployment is coming. The question is whether HR is at the table when it happens.

1. Map What AI Will Actually Replace in Your Function

"AI will change HR" is not a strategy. It's a headline.

Here's what a strategy sounds like: AI will handle 80% of Tier 1 HR queries within 18 months. That's IBM's number. It may not be your number exactly, but you need your own version of it — and you need it before job openings start vanishing in your own function and you're stuck explaining a headcount drop you didn't see coming.

Go task by task through every HR function you own. Which ones are information retrieval? Which are process execution? Which require genuine human judgment? That last category, at IBM, was 6%. Everything else was automatable — and increasingly, it's not just chatbots doing the automating. I've written before about what agentic AI actually looks like inside HR workflows, and it's a lot closer than "someday" for most of you.

Most CHROs haven't done this audit. They've read the trend reports. They've approved the pilots. But they haven't sat down and mapped which roles in their function become structurally unnecessary when the AI agent goes live.

Do that audit before the next deployment. Because when the CTO comes with the business case, you want to be the person who already has the workforce impact model, not the one scrambling to build it after the announcement is made.

2. Know Your AI Governance Accountability — and Take It

Grant Thornton's 2026 AI Impact Survey found that 78% of senior executives lack confidence their organization could pass an independent AI governance audit within 90 days. Three in four boards have approved major AI investments. Just over half have set clear governance expectations for that money.

That vacuum doesn't belong to IT. It doesn't belong to Legal. It belongs to HR.

Not because HR asked for it. Because HR is the only function with the mandate to govern how work gets done and how people are treated in the process.

Right now, AI governance accountability is floating. The CIO thinks the COO has it. The COO thinks the CIO has it. Legal is watching from a distance. And HR is sitting outside the room.

That needs to change before the next agent goes live, not after.

Go to your CEO and name it. "I want to own AI workforce governance — policy, accountability, audit process." If they push back, point them at the 78% number and ask which side of it they'd rather be on.

The Architect mindset means building the systems that govern how work happens. This is the most consequential system left on the table.

3. Redesign the Entry-Level Talent Pipeline Before AI Redesigns It for You

EY announced their Career Residency model in August 2026 — an 8-to-12-month paid pathway that replaces the traditional eight-week internship, built specifically because entry-level work no longer looks the way it did five years ago. Graduates come out at an elevated analyst level instead of the old bottom rung.

They built it because they had to. Twenty-two percent of organizations have already stopped hiring entry-level. By 2027, half expect to freeze or significantly reduce it.

Here's the twist worth sitting with: IBM — the same company automating 94% of Tier 1 HR — announced it's tripling its U.S. entry-level hiring in 2026. Not because the automation failed. Because the 6% that remained — the judgment calls, the exceptions, the human calls — turned out to be exactly where entry-level people build the instincts they'll need a decade later. Cut the on-ramp and you don't just have a hiring problem. You have a leadership bench problem that shows up ten years from now, when you look around for your next CHRO and find the pipeline empty.

You cannot simply inherit the old entry-level model and assume it still works. What does mastery look like now? How do you build judgment in people who aren't executing the work the old way? EY answered that question for themselves. You need your own answer, and it doesn't happen overnight. Start now.

4. Build a Human Skills Taxonomy for Every Role in Your Organization

The 6% of HR tasks IBM's AI couldn't handle — the judgment calls, the difficult conversations, the situations where context matters — that's your starting point.

What's in your 6%? What are the tasks in every department, not just HR, that genuinely require human capability? Empathy. Ethical reasoning. Navigating ambiguity. Building trust in a room where the stakes are real. Reading a situation that doesn't fit any playbook.

Every role in your organization has a version of this list. Most organizations have never written it down. They assume it. They value it intuitively. But they can't articulate it specifically enough to hire for it, develop it, or protect it when the next automation wave arrives.

Build that taxonomy. Audit every role against it. This work becomes the foundation of your future skills architecture, your performance criteria, your succession model. It's what lets you give a credible answer when your board asks what humans will be doing in five years.

The Scientist mindset means diagnosing with data. This is the diagnostic work your function has never had a reason to do before. Now it does.

5. Set the AI Use Policy Before Employees Set It for You

Employees are not waiting for permission. A Forbes Tech Council piece from August 2026 lays out the pattern plainly: large-scale surveys across the U.S. and U.K. put unapproved AI use somewhere between roughly half and two-thirds of employees, with senior leaders — the people writing the policies — often the most likely to work around them.

This is not primarily a compliance issue. It's a cultural signal. People are moving faster than the policy function can keep up with.

If you don't have a clear AI use policy, what you have is a workforce making risk decisions that belong to the organization — data handling, client confidentiality, accuracy standards, output ownership — made at the individual level, today, without governance.

Set the policy. Make it practical, not prohibitive — one employees actually use as a guide rather than route around. Make it fast enough to not be irrelevant by the time it's published. Then revisit it every six months, because the technology moves faster than your annual review cycle ever will.

The Choice in Front of You

CHROs who do these five things become indispensable to the AI transition. Not because they slowed it down. Because they made it work — with the workforce impact understood, the governance clear, the talent pipeline redesigned, the human skills identified, and the policy set before employees had to improvise one.

The ones who wait become the cautionary tale. You've seen the content — the function that was supposed to architect the workforce of the future, caught flat-footed by the thing they were supposed to lead.

That's not a technology failure. It's a preparation failure.

I wrote FutHRist because I believe the CHRO seat is being redefined right now, and the leaders who own it will be the ones building range across five mindsets: Architect, Engineer, Scientist, COO, and Coach. These five actions are the practical expression of that range.

The book launches soon. If you want to build the capability to lead this instead of react to it, start with the FutHRist Self-Assessment at trentcotton.com/futhrist-self-assessment.

Trent Cotton is a 20-year HR executive, author of FutHRist and Sprint Recruiting*, and host of The Human Capitalist podcast.*

Frequently Asked Questions

What percentage of HR tasks can AI actually automate?

At IBM, AI now contains 94% of routine HR queries — everything from vacation requests to pay statement questions. The remaining 6% requires human judgment: exceptions, ethical calls, and situations that don't fit a script. That 6% isn't a rounding error — it's where your future leadership bench gets built.

Who should own AI governance in an organization?

Right now, mostly nobody — 78% of senior executives say they couldn't pass an independent AI governance audit within 90 days. HR has the strongest claim to ownership because it already governs how work gets done and how people are treated, but that mandate has to be claimed, not assumed.

Will AI eliminate entry-level jobs?

It's already reshaping them. Roughly one in five organizations has stopped hiring entry-level, and half expect to freeze or cut it further by 2027. But companies like EY and IBM are responding by redesigning the entry-level path rather than eliminating it — because the judgment AI can't replicate has to be built somewhere.

What should an AI use policy actually cover?

Data handling, client confidentiality, accuracy standards, output ownership, and performance attribution — the risk decisions employees are currently making on their own without one. The policy needs to be practical enough that people use it as a guide, not something they route around.

How is the CHRO role changing because of AI?

The CHRO seat is shifting from a compliance and process function toward an architecture function — building the systems that govern how AI and people work together, redesigning talent pipelines, and defining what "human skills" mean role by role. The CHROs who shape that now will be indispensable; the ones who wait for it to happen to them won't be.