Insights
Oct 9, 2026

HR Can’t Lead an AI Transformation It Hasn’t Lived Through

HR is being handed the keys to AI transformation, yet lacks the literacy to lead it. The way forward starts with transforming its own work first.

Posted by:
Manuel Fortoul
Guest Author

Somewhere along the way, our whole profession was handed the keys to the AI transformation.

Nobody checked whether we knew how to drive.

I have spent nearly two decades inside this profession, and I like it enough to want it to survive its own mandate.

I have sat in both rooms. In one, the HR team automated its own work first, learned where the human belongs by doing rather than decreeing, and earned the right to lead everyone else through the same passage. In the other, the team is writing policies for tools nobody on it has ever operated.

So consider this an insider’s self-critique: we are being asked to write the policies, govern the tools, and redesign the workforce around a technology most of us have never seriously used.

The numbers are blunt. Culture Amp’s second annual AI in HR survey, published this July, found that 47% of HR leaders now claim ownership of their organization’s AI strategy; yet only 24% say they are comfortable letting AI act autonomously inside HR operations. Belief that AI will significantly improve how work gets done fell nine points, and confidence that generative AI will improve the perceived value of HR expertise fell from 47% to 38%. We are claiming the mandate while quietly losing faith that we can carry it.

The literacy gap has three faces

The first face is governing tools we cannot evaluate. Consider hiring, the highest-stakes thing most HR teams do. Employers now deploy AI screening tools that score and rank candidates at scale, and HR is expected to govern them; yet in Mobley v. Workday, a federal case filed in 2023, the plaintiff alleges he was rejected from more than 100 jobs through the same screening platform without landing a single interview, and in June of this year a federal judge refused to dismiss the case, allowing discrimination claims to proceed against the vendor itself. The old defense, that the employer made the final decision, is losing its shield. Layer on New York City’s Local Law 144, with its bias-audit and candidate-notice requirements, plus new AI employment rules in Illinois and California, and then ask the simple question: how many of the HR teams being asked to sign off on these tools could explain, in plain language, how the model scores a candidate?

Governance without literacy is compliance theater.

The second face is how we actually use the stuff: as a very smart intern. Culture Amp found that HR professionals spend their AI engagement on content creation, brainstorming, and synthesis, while only 39% have moved AI into HR operations automation. As Culture Amp CEO Caroline Rawlinson put it, “nearly every practitioner is using AI, but most are still using it the way they’d use a very smart intern.” Drafting job posts with a chatbot is not transformation; it is task-level tinkering.

The third is quieter, and worse. We cannot assess what we cannot do ourselves. A Talogy study published in August found that 78% of hiring managers face real challenges assessing AI skills in candidates, and only 38% feel very prepared to adapt job descriptions and career paths. We are being asked to hire for, develop, and plan around capabilities we have not built in ourselves. Protiviti’s AI Pulse survey found that on whether their organizations’ job designs are AI-ready, 13% of CHROs strongly agree, versus 28% of the broader C-suite. HR is not just behind the average: on every readiness question asked, it is the outlier.

Velocity is making it worse

This gap would be manageable if the technology were standing still. It is not; ManpowerGroup titled its July research “AI Adoption Is Outpacing Leadership Readiness,” and the title is the whole story. Every quarter, the models get more capable, the regulation gets more demanding, and the distance grows between what HR is asked to govern and what HR understands. The technology is not waiting for us to catch up.

You cannot close that distance with a training deck.

The way out is to go first

I have some sympathy for the teams stuck in this gap, because I have been that team. While running People at a global software company, we pointed our first serious AI experiment at the least flattering target available: our own workforce reporting. I built it myself, and the build wasn’t hard. What delayed deployment was simpler and more embarrassing: I was afraid something would fail in a way I couldn’t see, and that the exposure would poison every automation project that came after. We wired no-code integrations into our HRIS and watched the manual work quietly evaporate. The funniest part came after launch. For weeks, people kept asking me to double-check the data, certain the new system had skewed it. Then the tone of the questions shifted, from “is this right?” to “what are we going to do about this?” The tool wasn’t broken. What it was reporting on was. It was plumbing, not a moonshot, but plumbing teaches you things keynotes do not.

The technology was never the hard part. The data was, and cleaning it up forced us to confront years of accumulated inconsistency in how we recorded the basics. Adoption turned out to be a change-management problem in a technology costume: people trusted the manual process because they could see every step, so the new one had to earn that trust one accurate report at a time. And automating our own work showed us exactly where the human had to stay in the loop: in the judgment calls and in the accountability for whatever the data pointed to. You do not learn where the human belongs by theorizing about it. You learn it by watching the machine get something subtly wrong and catching it.

That experience left us with a method, and it is disarmingly simple: take any role and inventory it as a bundle of tasks, not a job title; sort the tasks into two piles: the ones AI can absorb and the ones where the human is the bottleneck because the work runs on judgment, relationships, or accountability; then redesign the role around the second pile and reskill toward it.

The unit of analysis is the task, never the headcount.

Companies that start with headcount end up doing layoffs followed by quiet rehiring, while companies that start with tasks end up with a workforce plan.

The profession’s choice

The encouraging news is that some People leaders have already figured this out. IBM’s CHRO Nickle LaMoreaux turned the company’s people function into what she calls an “AI-first” HR organization; its AskHR agent now handles about 94% of employee requests across more than 11.5 million annual transactions, and HR operational costs fell 40% over four years. She is also honest that the road was rocky, with employee satisfaction toward HR initially cratering before climbing to record highs, and that honesty is the credential. Atlassian, Moderna, and Lumen have each since put their head of HR in charge of company-wide AI transformation. As Fast Company put it, “The hard part of AI is not deploying the technology. It is reimagining the work and the workforce that will be doing it.” Boards are starting to understand that the person who should lead that reimagining is the one who has already done it to their own function.

So here is the choice. We can keep writing policies for tools we have never used and governing vendors we cannot evaluate, while our credibility quietly erodes. Or we can do the uncomfortable thing first: turn the transformation inward, automate our own work, and earn the right to lead everyone else through the same passage. As Rawlinson put it, “The question isn’t whether HR is ready. It’s whether leaders are willing to go there first.” The next generation of People leaders will be the ones who ran the experiment on themselves first.

The rest will be policing a transformation they never understood.

Manuel Fortoul has spent nearly two decades as a coach, consultant, and People leader, most recently as Head of People at a global SaaS company. He advises and writes on AI, talent, and organizational design, drawing on hands-on AI implementation inside the People function.

---

Sources

‍

Related posts

Browse all posts

We Keep Getting Surprised. And It’s Fantastic.

A first-hand log of how fast AI changed the way we work at Misfit Labs, and why we never stop experimenting.

Insights
October 6, 2026

Intelligence Over Headcount

AI is changing organizational capacity. The smartest companies optimize human and machine intelligence rather than headcount.

Insights
October 1, 2026