| This article references regulatory context including the EU AI Act’s Article 4 AI literacy obligation and ICO draft guidance on automated decision-making in recruitment. This is regulatory signal and workforce direction, not legal advice. For decisions that touch employment law, data protection or AI compliance, consult a qualified UK specialist. |
Last month I sat with an HR Director at a 110-person professional services firm in Birmingham. Her CEO had bought an AI CV-screening tool at a conference. She found out when the first rejected candidate complained about being filtered out by “a robot.” The tool had been running for six weeks. She could not explain how it worked, what data it used, or who had signed it off. She was the person on the hook.
She is not alone, and she is not behind. She is where most UK SME HR teams are right now.
AI is being bought by CEOs and dropped into HR’s lap. It is being built into recruitment tools, performance platforms and learning systems HR already uses. And when something goes wrong, it is HR who gets the call, not the vendor.
If AI training for HR teams is not already on your 2026 plan, this post is the argument for why it needs to be, and what to actually learn.
HR needs training before the rest of the business because HR already owns the systems AI is being bolted onto.
The CIPD surveyed over 1,300 senior leaders and HR professionals between 14 January and 3 February 2026. The headline is uncomfortable: HR capability is not keeping pace with AI. The skills HR already has, workforce planning, job design and employee engagement, are the ones that matter most for responsible AI adoption. Confidence in applying those skills to AI contexts is low.
The same research found a direct link to performance. In organisations where HR leaders held higher confidence in these skills, close to nine in ten reported that AI had improved worker job performance. Where HR confidence was lower, AI landed harder and delivered less.
The order matters. If a line manager starts using AI to write performance reviews before HR has agreed what good looks like, HR spends the next year cleaning it up. If a hiring team uses an AI screening tool before HR has worked out what “meaningful human involvement” means in practice, HR carries the regulatory risk when the ICO comes asking. HR going first is not about seniority. It is about who designs the system everyone else is trained to use.
I covered the broader workforce picture in a previous post on AI skills and the future of the UK SME workforce. This one is narrower and goes further. What HR itself needs to learn.
Four, in order: how the tools work, where human judgement must stay, how to write and enforce an AI policy, and how to challenge a vendor.
1. How the tools actually work. Not a degree in machine learning. The practical version. What a large language model is doing when it writes a job description. What an AI screening tool is actually scoring. Why the same CV returns different outputs on different days. Why the model will sometimes invent information and sound confident about it.
This is the base layer. Without it, nothing else lands. An HR team that cannot describe what their tool is doing cannot defend it to a candidate, a tribunal, or the regulator.
2. Where human judgement must stay. The ICO’s March 2026 report on automated decision-making in recruitment made a sharp point. Most employers told the ICO they were using AI tools for “decision support.” Evidence from engagement with more than 30 employers showed that in practice, decisions were often solely automated, with no meaningful human involvement. The hiring manager was rubber-stamping the tool’s output.
Under the Data (Use and Access) Act 2025, in force since 5 February 2026, the UK framework for automated decision-making has shifted. Article 22 of the UK GDPR has been replaced by Articles 22A to 22D, reframing automated decisions as a right of challenge with safeguards rather than a general prohibition. The ICO’s draft guidance, consulted on until 29 May 2026, is signalling what “meaningful human involvement” needs to look like in practice.
HR needs to know how to spot the difference between a tool that supports a decision and one that is making it. That is not a legal question. It is an operational one, and it belongs on the HR training plan.
3. How to write and enforce an AI policy. AI policy is HR’s territory. Not IT, not legal, not the CEO. HR decides what employees can and cannot put into a public AI tool, what the consequences are for breaching the policy, and how to help managers apply it fairly. Too many SMEs have no policy at all, and the ones that do often have one HR cannot explain.
I covered the policy itself in detail in AI policy for employees. The training question is whether HR can actually write one, explain it in a team meeting, and enforce it consistently when a star performer breaches it. That is a learnable skill. Nobody is born with it.
4. How to challenge a vendor. Most AI tools reach SME HR teams through a sales conversation, not a procurement process. HR needs the questions. What data trained this model? How often is it retrained? What is the error rate across different demographic groups? Can you provide the Data Protection Impact Assessment? What is your position on Articles 22A to 22D? What happens if we need to delete a candidate’s data?
A vendor who cannot answer these questions clearly is not a vendor to buy from. HR teams who have never been taught to ask them tend to sign anyway.
Different roles need different depth. One training plan does not fit a Head of People, an HR Business Partner and an HR Coordinator.
The EU AI Act’s Article 4 obligation, in force since 2 February 2025, requires employers who deploy AI systems to ensure a sufficient level of AI literacy among staff. The European Commission’s May 2025 Questions and Answers clarified that literacy is role-based, not uniform. That principle holds even if the SME is UK-only, because the practical shape of it is the same.
A rough split that works for an SME HR team of three to eight people:
Head of People or HR Director. The deepest layer. Vendor challenge, policy ownership, regulatory signal-reading, workforce planning for an AI-affected organisation. This is the person who sits in the room when the CEO buys the next tool. If they cannot push back with specifics, the tool gets bought anyway.
HR Business Partner or Senior HR Manager. Enough to coach line managers, run training sessions, and make judgement calls on individual cases. Can they explain the AI screening tool to a hiring manager? Can they spot a performance review that was AI-generated and feels off? Can they tell when a line manager is over-relying on an AI output?
HR Coordinator or HR Administrator. The practical day-to-day. How to use the tools HR itself has adopted. What data never goes into a public model. When to escalate. How to log an issue.
If you only have one HR person, they wear all three layers. The learning plan sequences by layer, not by job title.
Start with free government and professional-body resources, add one internal champion, and build a 90-day plan around real HR work.
The CIPD has published an AI skills planning guide for people professionals, developed through the Innovate UK BridgeAI programme. It is anchored in people practice, not technology. First stop for anyone with a CIPD membership.
Skills England’s AI Skills Framework, Adoption Pathway Model and Employer Checklist, published through gov.uk in late 2025, are open access. The framework is sector-neutral but the HR-relevant subset is clear.
The AI Skills Boost programme, announced on 28 January 2026, aims to upskill 10 million UK workers in AI skills by 2030 through a government and industry partnership. It is free at the point of use and good for the base layer, though not HR-specific.
Three practical moves have worked for the SME HR teams I have advised:
Appoint one AI champion inside HR. Someone who gets genuinely interested, is given the time to go deep, and becomes the person others ask. One champion in an eight-person HR team is enough.
Build the plan around real HR work. A quarter spent getting good at using AI for job description drafting, interview scheduling and policy writing beats a year of generic AI literacy modules. HR learns by doing HR work with AI, not by watching slide decks.
Set a six-month check-in built around outcomes. The right question is not “did everyone complete the training.” The right question is “can our HR team now write an AI policy, challenge a vendor, and explain to a candidate how an AI tool made a decision about them.” If the answer is no, the plan did not work. Change it.
Once HR is trained, the roadmap for rolling AI out across the wider hiring process is covered in implementing AI in hiring.
An HR team that can write the policy, challenge the vendor, train the line manager, and explain the system to the candidate.
Not a certificate wall. Not a maturity model. Four tests:
Can the Head of People sit opposite an AI vendor and ask the five questions that matter? Can the HR Business Partner coach a hiring manager through using an AI screening tool without rubber-stamping it? Can the HR Coordinator explain to a new hire what data goes into the company’s AI tools and what does not? Can HR, as a function, explain to the ICO or to a candidate how an automated decision was made?
That is the bar. It is reachable in six months for an SME HR team that treats this as real work rather than a training compliance exercise.
The HR Director I started this post with has started. She has a policy draft, a vendor review scheduled, and one AI champion assigned on her team. She is still worried about the next tool her CEO brings home from the next conference. The difference now is she has a plan for it.
That is what AI training for HR teams actually looks like. Not a syllabus. A set of real-world capabilities that let HR do the job AI has just made more complicated.
In most cases, yes. HR designs and enforces the systems, policies, screening and performance management that the rest of the workforce uses. If HR is not trained first, the organisation ends up with inconsistent AI practice that HR then has to retro-fit policy around. The CIPD’s 2026 research signals that HR capability is a strong determinant of whether AI improves worker performance.
Article 4, in force since 2 February 2025, requires deployers of AI systems to ensure a sufficient level of AI literacy among staff. It applies to UK-based employers who deploy EU-provided AI systems or whose AI systems affect people in the EU. Enforcement begins on 2 August 2026 under European Commission guidance. UK employers hiring EU candidates or using EU-built AI tools should treat it as a planning baseline.
The Data (Use and Access) Act 2025, in force since 5 February 2026, replaced UK GDPR Article 22 with Articles 22A to 22D. The ICO’s March 2026 draft guidance, consulted on until 29 May 2026, is signalling how it expects employers to apply the new framework in recruitment specifically. Final guidance was expected in summer 2026.
The ICO’s working position, from its March 2026 recruitment report, is that human involvement is not meaningful if the reviewer is simply approving the tool’s output without the ability or training to challenge it. HR teams need to train hiring managers in how to actually review an AI output, including when to override it.
No. Neither the EU AI Act nor UK law requires one. What works for SME HR teams is appointing one internal AI champion inside the function, someone who goes deep, keeps the team updated, and becomes the point of escalation when a judgement call is needed.
| Sabiha is a Talent Acquisition Director and Speaker with 16+ years of international hiring experience across the UK, Dubai, South Africa and Malaysia. She has advised 300+ UK SMEs on AI-enabled hiring and retention, and was shortlisted for Best Career Coach UK by the Career Development Institute. Her forthcoming book, How to Use AI to Win Talent and Retain People (Trotman), publishes Autumn 2026. |
Global Talent. Ethical AI. Strategic Hiring. Sustainable Retention.
