Quick Answer: AI agents for HR are software systems that carry out multi-step HR work across tools, such as answering policy questions, keeping records in step and preparing casework. They work best on the admin around decisions about people. Decisions with a significant effect on someone, such as hiring, dismissal or pay, should stay with a person, and UK law now sets safeguards for any made without meaningful human involvement. The best starting list comes from the HR team’s own week.
Only 39 per cent of organisations have put AI to work in HR, against 62 per cent using it somewhere in the business, according to SHRM’s survey of more than 1,900 HR professionals, published in 2026 by the US HR professional body. Where HR does use AI, recruiting leads at 27 per cent, the area where decisions about people carry the highest stakes. Most guides to AI agents for HR start from a catalogue of use cases, each tied to a product, describing AI agents for human resources in general rather than one team’s work. A better list starts with what fills the HR team’s own week, sorted by one question: is this admin around a decision, or the decision itself? The same test applies to AI agents across the wider business, but HR holds more personal data and more consequential decisions than most teams.
What HR Work Actually Looks Like
HR serves two groups at once: the employees whose working lives it administers and the managers who apply its policies. Any useful list of AI agents for HR starts from the six kinds of work that fill its week.
- Answering questions. Holiday, sickness, family leave and pay dates, asked every week in different words, with answers that often depend on contract type.
- Keeping records in step. Contract changes, hours and personal details have to match across payroll, the HR system and the rota.
- Running people processes. Recruitment admin, probation reviews and annual pay and appraisal cycles, all on deadlines someone has to track.
- Casework. Absence, disciplinary, grievance and flexible working cases are documented, time-limited and confidential.
- Advising managers. Managers rarely ask what a policy says. They ask how it applies to this person, this week.
- Reporting. Headcount, absence and turnover figures for leadership, often rebuilt by hand.
Only some of this appears in a job description. The rest sits in inboxes, spreadsheets and the habits of people who know which manager needs a reminder before a probation review slips.
Why HR Work Is Hard to See
Most HR teams can list their processes. Fewer can show where the hours go, because much of the work happens between systems. An Ipsos survey for Eagle Hill Consulting, a firm that sells HR transformation work, asked 200 HR professionals at US organisations with revenue above $100 million about this in March 2026. The sample is small, but 91 per cent used two or more systems for a single task, 78 per cent regularly entered the same information into more than one system, and 68 per cent relied on workarounds such as spreadsheets and shared drives.
Those workarounds are close to invisible from outside, which matters for AI agents for HR, because nobody automates work they cannot see. A job description says “advise managers on policy”, not “copy one contract change into three systems”. That knowledge sits in the unwritten side of the job, and confidentiality hides casework further still.
HR is also often missing when AI plans are made: the SHRM research found that 52 per cent of organisations do not involve HR in AI strategy. The reliable way to find the real list is asking the people who do the work, and the managers who depend on them.
Where AI Agents for HR Fit
Three kinds of work suit an agent, and all three sit around decisions rather than inside them. Agents earn their place on the parts of the week that depend on someone remembering, re-keying or chasing.
Answering the Same Question in Different Words
An agent can answer managers’ and employees’ questions from current policy, show the clause it relied on and pass anything unusual to an adviser, so the answer no longer depends on who opens the email. A record of the questions that keep coming back also shows where a policy is unclear.
Keeping Records in Step Across Systems
When one change has to land in payroll, the HR system and the rota, an agent can draft it once and flag mismatches before they reach a payslip. Calculations that follow a fixed formula are a different case, and it is worth knowing where fixed rules do the job better before choosing. Many useful AI agents for HR end up working alongside it.
Preparing the Ground for Decisions
Much of HR’s effort goes into getting a decision ready rather than making it. An agent can build a case timeline from records, draft a letter for an adviser to review, summarise applications against agreed criteria and book a hearing or interview. Used this way, AI agents for HR prepare and a person decides. For some decisions, the law now spells out what happens when that person is missing.
What Should Stay With a Person
The Data (Use and Access) Act 2025 replaced the UK GDPR’s old rules on automated decisions from 5 February 2026. A significant decision has a legal or similarly significant effect on someone, and it is solely automated if no person is meaningfully involved. Where it is, the employer must tell the person and let them make representations, obtain human intervention and contest it. Decisions resting even partly on health or other special category data cannot be solely automated except with explicit consent or under narrow legal conditions. Specific cases need legal advice, but the more a decision changes someone’s working life, the more a person should be part of it.
What the law permits and what staff will accept are separate tests. In polling for the TUC, carried out by YouGov in April 2024, 77 per cent of working UK adults opposed AI making hiring decisions and 86 per cent opposed it making firing decisions. HR depends on trust, and AI agents for HR that leave people feeling processed will cost more than they save.
| Kind of work | Health or special category data? | Significant decision? | What an agent can do | Where a person signs off |
| Policy questions | Rarely | No | Answer from current policy and route exceptions | Any exception to policy |
| Updating records across systems | Sometimes, such as absence | No | Draft changes and flag mismatches | Changes to pay or contract |
| Recruitment shortlisting | Can arise | Yes, if rejections are automated | Summarise applications against agreed criteria | Every shortlist and rejection |
| Sickness absence and adjustments | Yes | Often | Gather records and schedule meetings | The decision and the conversation |
| Disciplinary and grievance cases | Sometimes | Yes | Build the timeline and draft letters for review | Every finding and outcome |
| Pay and promotion | Rarely | Yes | Prepare data and comparisons | The decision itself |
One HR Team, Five Pieces of Work
Consider a hotel and restaurant group with about 600 staff across 14 sites, many on irregular hours or seasonal contracts, and an HR team of six. Before looking at AI agents for HR, the group asked its HR team and site managers what filled their week. Five pieces of work came back.
| Piece of work | Raised by | What it involves | Best handled by |
| Holiday and sickness questions from site managers | HR advisers | The same questions weekly, with answers that depend on contract type | An agent answering from policy, exceptions to an adviser |
| Holiday entitlement for irregular-hours staff | Payroll officer | Worked out from hours each pay period, keyed into two systems | Rules-based automation |
| Seasonal recruitment admin | Site managers | Interview scheduling and candidate updates in the spring peak | An agent, with site managers confirming slots |
| Shortlisting seasonal applicants | Head of HR | High application volumes every spring | An agent summarises; a site manager decides every shortlist |
| Adjustments for staff returning from long-term sickness | HR advisers | Health information, occupational health advice and a conversation | Stays with people |
Four of the five came from people doing the work, not the head of HR. The one item leadership raised, shortlisting, ended up with the tightest human sign-off, because it sits closest to a significant decision. From a distance, the decision looked like the obvious thing to automate. From inside the team, the admin around it was where the time went.
Making AI Agents for HR Pay Off
A few habits separate agents that stay in use from those switched off after a pilot.
- Build the list from HR’s own week. Include the managers who rely on HR, where many repeat questions start.
- Write down the line between admin and decision. Each agent needs a named person who signs off, plus the four safeguards wherever a significant decision could be automated.
- Keep health data away from automated decisions. They need a clear lawful basis and access limited to what each agent needs.
- Tell staff what is automated and where a person decides. In the TUC polling, 69 per cent wanted staff consulted first. HR also fields “will this replace us?” for every department, and a clear account of what AI agents for HR will change answers it better than reassurance.
- Take a baseline and measure. SHRM found 56 per cent of HR functions do not formally measure their AI investments, and hours saved mean little without a baseline taken before launch.
- Hold agents centrally. HR’s agents should belong to the organisation, not one adviser’s account. HR usually writes the policy on AI tools staff adopt on their own, so its own agents should meet it and keep running when an adviser moves on.
Where GrowthNation Fits
GrowthNation starts from the work rather than a catalogue. Everyone in the HR team, and the managers who rely on it, is interviewed by an AI agent, by voice or text and in confidence, about how they work and what slows them down. Agents are built straight from those interviews and served centrally, so AI agents for HR built this way stay with the organisation. Once they are in use, hours saved are tracked per agent and rolled up across the team. Some HR work will suit an existing HR platform or simple rules-based automation, and decisions about people stay with people.
Map Your HR Team’s Week Before You Choose a Tool
The quickest way to see what belongs on HR’s list is to ask the team. A walkthrough maps a single team, shows you the scorecard and hands you a working agent built from your own week, not a catalogue.
Book a walkthrough: one team, one scorecard, one working agent. No deck.
Join the GrowthNation waitlist to be among the next organisations to start.
Frequently Asked Questions
What are AI agents for HR?
AI agents for HR are software systems that carry out multi-step HR work with limited supervision, such as answering policy questions, updating records across systems and preparing case files. Unlike chatbots, they act across several tools. They suit the admin around people decisions, while decisions with a significant effect on someone stay with a person.
What HR tasks can AI agents handle?
AI agents in HR suit frequent, rule-bound work: answering leave and sickness questions from current policy, keeping payroll and HR records consistent, scheduling interviews and drafting letters or case timelines for review. Fixed calculations such as holiday accrual often suit rules-based automation. Hiring, dismissal, pay and adjustment decisions should stay with a person.
Can AI make hiring decisions in the UK?
Since 5 February 2026, the Data (Use and Access) Act 2025 has allowed solely automated significant decisions, including hiring decisions, but only with safeguards. The candidate must be told, and must be able to make representations, obtain human intervention and contest the decision. Stricter limits apply where health or other special category data is involved.
Should AI agents use employee health data?
Only with care. Under UK data protection law, a significant decision based even partly on health or other special category data cannot be solely automated without explicit consent or a narrow legal condition. In practice, sickness absence and adjustments are best handled by people, with AI agents for HR limited to scheduling and paperwork.
Do employees need to be told when AI is used in HR decisions?
Where a significant decision about someone is solely automated, UK law requires the employer to tell them and offer ways to make representations, obtain human intervention and contest it. Beyond that minimum, telling staff what is automated and where a person decides is good practice, and TUC polling from 2024 found 69 per cent of working UK adults wanted staff consulted first.
How should an HR team choose its first AI agent?
Start with work that is frequent, follows clear rules and carries low stakes for any individual, such as repeat policy questions or keeping records in step. The best candidates come from the HR team and its managers, not a vendor catalogue. Take a baseline first, and leave decisions about people out of the first round of AI agents for HR.
