AI in HR
What Is the Future of AI in HR? Five Shifts Already Underway, and What the CHRO Does About Them.
AI is taking HR's transactional work, recruiting first. What comes next is a function that designs how people and agents work together, if HR leads the redesign instead of receiving it.
Ask most HR teams where AI fits and the answer so far has been recruiting. Résumé screening, interview scheduling, a chatbot for benefits questions: AI went in where the volume was highest and the friction most obvious.
Meanwhile the company's AI strategy was being written somewhere else, usually by IT, and now the executive team expects HR to run a workforce made of people and software agents. The distance between those two sentences is the future of AI in HR, and it has already arrived.
The useful answer is less about which tasks a model can do, since that list grows every quarter, and more about what the function becomes once the tasks are gone.
Where AI in HR actually stands in 2026.
Start with where things are, because most of the pages on this subject still quote 2022. SHRM's State of AI in HR 2026, published in March from a survey of more than 1,900 HR professionals, found that 39% of organizations have implemented AI in their HR function and another 7% plan to this year.
Thirty-one percent have no plans at all. The gap by size is wide: 60% of extra-large organizations have adopted AI in HR, against about one in three small ones. And AI is further along everywhere else in the business.
Sixty-two percent of organizations use it somewhere, and 23% use it in the company but not in HR.
Where it is used is narrow. Recruiting leads at 27% of organizations, then HR technology at 21% and learning and development at 17%, and seven of the ten most common use cases are in recruiting.
SHRM's lead HR technology analyst, Kenny Pyle, explained it simply: organizations start where the volume is highest and the friction most obvious, and in HR that is almost always recruiting.
Two more numbers from the same study matter more than the adoption rate. Fifty-six percent of HR functions do not formally measure whether their AI investments are working, and 16% use return on investment as a yardstick.
And 52% of organizations, SHRM found, do not involve HR in the company's AI strategy or vision at all. The function that owns the workforce is, in half of companies, outside the room where the workforce's biggest change is being decided.
What the future of AI in HR is being built on: where adoption stands in 2026
Adoption is real, narrow and unmeasured, and half of HR functions are outside the AI strategy.
Source: SHRM, The State of AI in HR 2026 (March 2026)
Even where adoption is close to universal, the value is not. In Gartner's survey of 110 heads of HR at the end of 2025, 95% said their organizations had implemented AI in some form over the past year, and one in five had seen significant or transformational value from it (Gartner, July 2026).
Adoption is real. It is also shallow. HR has automated the front door, left the rest of the house alone, and in most cases has not checked whether the front door works.
Will AI replace HR?
No, not the function. It will take a large share of the function's work, and the people who do that work will move: into roles the function did not have, into the judgment work the agent cannot do, or out. That is a different sentence from "AI augments HR", and the data behind it is more useful than the reassurance.
In SHRM's 2026 study, 7% of organizations reported job displacement because of AI. Against that, 39% reported shifts in job responsibilities, 24% reported new roles created and 57% reported more upskilling and reskilling.
Responsibilities were more than five times as likely to change as to disappear. The HR professionals using AI report the gains themselves: 87% say it has improved their efficiency, 75% the quality of their work and 70% their creativity.
What AI did to HR jobs in 2026, by share of organizations reporting it
The work is being reshaped far faster than the jobs are being cut.
Source: SHRM, The State of AI in HR 2026 (March 2026)
Those are this year's numbers, and the transactional share of HR's work is large, so expect the displacement figure to rise. The shape will hold.
In Rewired Leadership I argue that AI will not replace leaders; it will expose them, by making every outdated habit visible, fast. The same is true of the function. AI does not replace HR.
It exposes which parts of HR were administration all along, and administration is not a reason for a function to exist. What remains, and gets more valuable, is the work only people can do: judgment about people in situations where the answer is not in the data.
Five shifts that define the future of AI in HR.
The future of AI in HR is not a forecast. Each of the five shifts below is already underway, with the evidence to show it, and each does the same two things: it moves a piece of execution off HR's desk and puts a design decision on it.
| The shift | What AI takes | What HR keeps | The metric |
|---|---|---|---|
| 1. Assistants to agents | Screening, scheduling, answers, filing, the workflow itself | Deciding which work stays human | Share of human time freed |
| 2. Activity to judgment | Counting output | Defining and rating the quality of decisions | Decision cycle time, delegation rate |
| 3. Job titles to capability clusters | Matching people to tasks | Designing the structure agents work inside | Talent mobility rate |
| 4. Retention to unlearning | Delivering training at scale | Deciding what people must stop doing | Time to unlearn |
| 5. Policy to governance and the emotional contract | Drafting, monitoring, flagging | Accountability, disclosure, trust | Employees who know how AI is used on them |
01From assistants to agents: HR stops storing records and starts running workflows
Every vendor roadmap says "agentic" this year, so it is worth saying what it means. An assistant drafts the job description when asked.
An agent screens the applicants, schedules the interviews, answers the candidate's question about the policy, files the change when the offer is accepted and starts on the next one, with a person checking the exceptions.
The HR platform stops being a system of record that people operate and becomes a system that does work.
Mercer's Global Talent Trends 2026, a survey of nearly 12,000 executives, HR leaders, investors and employees, found that 82% of C-suite executives say the future of the HR function lies in managing human talent and digital agents side by side (Mercer, February 2026).
What HR keeps is the sorting. The book's tool for it is the Judgment Audit: map every function's output into three columns. Execute, which the agent does. Synthesize, which the agent and a person do together. Judge, which stays with a person and gets elevated.
The hard column is the first, because filling it means settling which decisions stay with a person and which an agent may take inside the guardrails.
In the ninety-day plan at the end of the book, the audit runs in the first two weeks with a target of freeing a quarter of human time. HR runs it with the CTO, on the whole organization, and on itself first.
02From activity to judgment: the performance system gets rewritten
When an agent does the execution, a review form that counts tasks closed, cases handled and hours logged is scoring the machine, and paying the person to compete with it.
People notice what gets rated, and they protect that work even when the agent does it better, which is how AI rollouts stall without anyone quite saying why. Nothing reinforces old behavior faster than an outdated compensation system.
So the first of the three mandates I set out for the CHRO in the book is to shift the whole performance system to measuring the quality of human judgment, not the volume of human activity.
In practice that means deciding what the review measures once an agent does the execution: decisions made well inside the guardrails, decision cycle time, the quality of what gets escalated, and how much complexity a person removed.
Rating criteria are HR's own instrument, which makes this the one shift the function can start on tomorrow without asking anyone.
03From job titles to capability clusters: the structure changes
Agents do not read org charts. They create workflows that cut across functions, and a structure built from fixed job titles and reporting lines becomes the bottleneck.
Mercer found that 98% of executives plan organizational design changes over the next two years and 63% say they need to move toward skills-powered talent practices.
The book's version of that shift is to define jobs by capability rather than title, a "critical risk assessor" instead of a "senior analyst", so that people can be deployed across agent-augmented workflows as the work changes, and to track the talent mobility rate: the share of people who move between significant capability clusters within eighteen months.
A healthy organization moves more than a quarter of its high-potential talent a year. Structure follows the metric: federated, modular teams with decision rights set by the guardrails, which usually means fewer management layers.
Succession has to move too, because in most companies the successors were chosen against criteria written before AI.
04From retention to unlearning: learning and incentives
The old learning and development question was how to retain skills. The new one is how fast people can stop doing what an agent now does and learn the higher-judgment work.
Mercer's data gives the scale: 65% of executives expect 11% to 30% of their workforce to be redeployed or reskilled because of AI within two years, and 63% of employees say they would trade a 10% pay rise for the chance to upskill in AI and digital skills.
The book proposes three instruments. An explicit unlearning budget: funded time for high-value people to deliberately stop performing obsolete tasks and train on the judgment work.
A metric, time to unlearn: how quickly a skill is handed to an agent and the person is working at the next level up. And agent literacy for every employee, not just the HR team: prompting, bias, and responsible use.
Incentives follow the metric. A meaningful share of a senior leader's bonus should ride on their team's delegation rate and talent mobility, so that leaders are paid for what their people can do without them.
05From policy owner to governance and the emotional contract
Governance is the shift HR cannot delegate, and the one it is least ready for. In SHRM's study, 49% of organizations have a policy on employee AI use, and only a quarter of those think the policy is future-proof.
Among HR professionals in US states with workforce AI regulations, 57% are unaware of the rules, and of those who are aware, 12% have acted on them. SHRM counts nineteen of the most populous states with employment AI laws.
In Europe, the Digital Omnibus on AI, in force since 27 July 2026, moved the EU AI Act's obligations for high-risk employment and recruitment systems from August 2026 to 2 December 2027, as the law firm Lewis Silkin set out at the time: later than planned, but not far.
The quieter half is the emotional contract. Mercer found that 40% of employees are concerned about losing their job to AI, up from 28% in 2024, and 62% say leaders underestimate AI's emotional impact.
Nineteen percent of HR leaders include that impact in their digital implementation strategy. Over the same two years, the share of employees who say they are thriving at work fell from 66% to 44%.
You cannot hand people an agent-run workflow and expect them to use their judgment inside it while they are afraid of it, and fear is settled by how bad news and mistakes are handled, not by the policy.
The emotional side of AI in HR, 2024 against 2026, from Mercer's 2026 report
Fear went up, thriving went down, and executives got less sure of themselves.
Source: Mercer, Global Talent Trends 2026 (February 2026)
Who leads it: the CHRO as chief transformation agent.
Here is the question the five shifts all come back to. Each one needs someone to redesign a people system: roles, ratings, structure, incentives, rules. In half of organizations, HR is not involved in the company's AI strategy at all.
The C-suite's top people priority, in Mercer's survey, is redesigning work to incorporate AI; HR leaders' top priority is improving the employee experience.
IBM's Institute for Business Value, surveying 1,500 CHROs this year, found that 46% of organizations do not involve the CHRO when AI strategy is defined (IBM, September 2026). The function that owns the workforce is in danger of receiving the redesign, and then being asked to run it.
Deloitte's 2026 Global Human Capital Trends, drawn from more than 9,000 business and HR leaders in 89 countries, shows what that costs.
In Deloitte's companion research with C-suite leaders, most organizations, 59%, were taking a technology-focused approach to AI, and those organizations were 1.6 times more likely than human-centered ones to fail to see returns beyond what they expected (Deloitte, March 2026).
The report's verdict on the core functions, HR among them, is that they are increasingly too slow and siloed for what the business now demands.
So the CHRO's job changes more than anyone's. In Rewired Leadership I put it this way: the CHRO is no longer a strategic partner but the chief transformation agent, the executive owner of the organization's adaptive capacity. The CEO champions the vision and the CTO manages the technology.
It is the CHRO who redesigns the structure, the talent pipeline, the metrics and the incentives, and who coaches leaders through a job no mentor has done, because every technological investment ultimately fails or succeeds on the architecture of the people system.
I have spent thirty-five years advising executive teams, and the last decade working with dozens of CHROs on exactly this transition, and that lesson has not had an exception.
The ninety-day plan in the book gives the three roles their numbers. The CEO owns the delegation success rate, the share of delegated decisions that are not quietly pulled back. The CTO owns decision cycle time.
The CHRO owns the talent mobility rate, and the first task is to kill the job description: run the Judgment Audit and remove 40% of low-value execution by day sixty. The three meet daily for the ninety days. Not weekly.
Most CHROs I meet are being asked by their board whether the company is ready for AI, and are answering with an inventory of licenses and pilots.
The honest answer depends on the leaders and the people systems, which is why the first thing we do is a readiness check that scores the leaders, the talent processes and the performance system rather than the data.
A July 2026 Gartner survey of 297 CHROs found that 78% of leaders have no alignment on where AI creates value (Gartner, October 2026). Those 78% are the point of the exercise.
What the future of AI in HR looks like in practice.
Three examples, one illustrative and two from the book, show the shifts at the scale of a team, a leader and a function.
A recruiting team after the agent. Picture a team where the agent now screens every application, schedules every interview and answers candidates' questions. Recruiting is where AI in HR starts, in SHRM's data, so this is the common case.
The old measures were time to hire and requisitions closed, and the agent flattens both; every recruiter looks the same. The redesigned measure is the calls the agent cannot make: which finalist to back when the data is split, which offer to walk away from, when to tell a hiring manager the brief is wrong.
Time to hire still shows on the dashboard. It no longer shows on the review. This example is illustrative, not a client case.
The CLO and the thirty hours. From Rewired Leadership: a chief learning officer spent thirty hours a month correlating course completion, performance data and business outcomes by hand.
An agentic system took over the correlation and flagged the anomalies, such as the top-performing region quietly ignoring the mandatory sales training, and her review time fell to three hours on one dashboard.
The time went to challenging the design of the curriculum. She stopped being the reporter and became the architect of the learning strategy. That is shift four, at the scale of one leader.
The Global 100 CHRO. Also from the book, with the name changed: a CHRO at a Global 100 technology company was leading a multi-year leadership transformation across 80,000 employees to prepare the company for an agentic operating model.
The strategy was sound and the funding was there. Six months in, regional initiatives had stalled, and the cause was not resistance below but the executive committee above.
Budgets the CFO had already approved waited an average of forty-five days for the committee's review, because every member felt obliged to add value. Her diagnosis was that her leaders were drowning in the effort of being indispensable.
The fix was one habit, publicly unlearned: the CEO stopped reviewing how budgets were spent and reviewed only the clarity of the intent and the ethical guardrails. That freed more than five hundred hours of executive time and told the organization the trust was real. The CHRO led that. It was not an HR initiative.
How HR leaders prepare for the future of AI in HR.
AI transformation in HR runs in a different order from the technology rollout, and the first step decides whether the rest matters. This is how the work goes when I do it with a CHRO.
- Get into the AI strategy room. If HR is not shaping how AI gets applied to work, it is reacting to decisions already made elsewhere, which is how SHRM's Kenny Pyle described the risk. The entry ticket is a position on the people systems above, with numbers attached, not a request to be consulted.
- Run the Judgment Audit on HR's own work first. Aim to free a quarter of the team's time. A function that has not redesigned its own work has little standing to redesign everyone else's, and the procedure that defines the execute bucket is where the handoff to an agent gets written down.
- Rewrite what performance measures, and how AI-assisted work is credited. This is the one shift HR controls outright. If the answer to the second question is "the same as before", the system is scoring the machine.
- Build agent literacy for every employee, not just HR. Prompting, bias, when to override the agent and when not to. The book's checklist makes it one of the nine questions the CHRO answers every quarter.
- Set governance before the regulator does. Know which state rules apply, plan for the EU's December 2027 date, and write a disclosure rule for AI in people decisions: 78% of managers have used AI on feedback or reviews, and 16% of employees were told (Highwire, October 2026).
- Measure it, because 56% of HR functions do not. Time to unlearn, talent mobility rate, delegation rate, and the return on investment the board will ask about. What gets measured is what the function can defend.
That sequence is the design step of the talent and performance work we do: redesign performance, talent processes and structure so they reward the work people actually do now, piloted with one team before the business.
Pros and cons of AI in HR.
The case for AI in HR is strong and the risks are specific, so the honest list is short on both sides.
| Pros | Cons |
|---|---|
| Efficiency: 87% of HR professionals using AI told SHRM they saw an improvement | Bias at scale, and the audit duty that comes with it under state law and the EU AI Act |
| Work quality and creativity: 75% and 70% report improvements | Over-monitoring: the tools that answer questions can also watch people |
| Consistency: the same policy answer for every employee, at any hour | Undisclosed AI in people decisions: 78% of managers use it in reviews, 16% of employees are told |
| Throughput where the volume is highest, which is recruiting | Skills atrophy when people stop doing the work they are still rated on |
| Evidence for decisions that used to run on a manager's memory | The emotional cost: 40% of employees fear for their jobs, and most leaders underestimate it |
The disadvantages of AI in HR are not reasons to wait. Each one is a design decision HR has to make before the agent makes it by default.
Frequently asked questions.
Today, execution: screening, scheduling, answering routine questions, drafting, and flagging patterns in workforce data, with recruiting the most common use.
Next, the role shifts to HR designing the system that people and AI agents work inside: which work stays human, how it is rated, how the organization is structured and what gets rewarded.
Some, and it will reshape far more. In SHRM's 2026 study 7% of organizations reported job displacement from AI, while 39% reported shifted responsibilities and 24% reported new roles.
Transactional HR work is the most exposed. The judgment work, on people in ambiguous situations, becomes more valuable, and so do the people who do it.
An agent runs a whole workflow rather than answering one request. In recruiting, it screens applications, schedules interviews, answers candidates' questions and files the hire, escalating only the exceptions.
The human recruiter keeps the calls the agent cannot make: which finalist to back, which offer to walk away from, when the brief itself is wrong.
Narrowly. SHRM found recruiting leads at 27% of organizations, then HR technology at 21% and learning and development at 17%, and seven of the ten most common use cases are in recruiting.
Compliance and inclusion barely register. Only 39% of organizations have implemented AI in HR at all, against 62% using it somewhere in the business.
Bias at scale, and the legal duty to audit for it. Monitoring that drifts into surveillance. AI used in reviews and other people decisions without telling the people.
Skills that atrophy when the agent does the work people are still rated on. And fear, which 40% of employees now report and most leaders underestimate. Each is a design decision, not a reason to wait.
Get into the AI strategy conversation with a position, run the Judgment Audit on HR's own work first, rewrite what performance measures, build agent literacy for everyone, set governance and a disclosure rule before the regulators arrive, and measure the results.
If you are not sure where to start, an assessment of the leaders and the people systems is step one.
The short version.
The future of AI in HR is agents doing the execution and HR designing the system around them: the roles, the ratings, the structure, the incentives and the rules. Adoption today is real, narrow and unmeasured, and in half of companies HR is outside the AI strategy.
The function will not be replaced. It will be exposed, and then redesigned. The only question that matters is whether the CHRO designs it or inherits it, and the talent and performance work we do starts with that choice.