Articles

Don’t rely on AI at work: 5 HR decisions that still need human judgement

Artificial Intelligence can be brilliant for speeding up admin, but it’s risky to let it make (or heavily steer) people decisions. If you’re a UK employer using AI tools for recruitment, performance, or drafting HR documents, the key is simple: AI can assist — but humans must stay accountable.

Below are five common areas where over‑reliance on AI can increase legal and reputational risk, plus practical steps to stay compliant.

1) Shortlisting candidates: bias can creep in quietly

Many employers use CV screening or keyword tools to triage applications. Done well, it saves time — but it can also embed bias if the model is trained on historic patterns.

Practical safeguards

• Keep humans responsible for final shortlist decisions.
• Run regular “spot checks” on rejected applications to look for patterns that disadvantage protected groups.
• Be transparent: explain the use of AI in your recruitment privacy notice, and document it in a DPIA.

2) Drafting employment documents: UK wording matters

AI is useful for first drafts, but it can mix up UK rules with other jurisdictions or produce clauses that simply don’t hold up in a dispute.

Practical safeguards

• Use AI for structure and plain‑English rewriting — not as the final legal source.
• Have contracts, policies, and settlement agreements reviewed by a qualified employment law professional.

3) Dismissal and redundancy decisions: automated scoring is high risk

Productivity trackers and AI “rankings” can be tempting — especially during restructures. But relying on automated data to decide outcomes can lead to unfair dismissal and discrimination claims.

Practical safeguards

• Ensure decision‑making includes meaningful human review.
• Keep a clear audit trail of the rationale and evidence used.
• Give employees a fair chance to respond (a process, not an algorithm).

4) “Replacing older workers with AI”: avoid age discrimination traps

Automation may be commercially sensible, but if changes disproportionately affect older workers (or leadership messaging suggests that’s the goal) it can become evidence in an age discrimination claim.

Practical safeguards

• Use objective selection criteria for role changes and redundancies.
• Review impact by age group and mitigate where needed.
• Be careful in internal and external communications about “AI replacing people”.

5) Using AI as “employment advice”: confident doesn’t mean correct

AI tools can sound authoritative while being wrong, out of date, or missing context. That can quickly turn into costly mistakes.

Practical safeguards

• Use AI to generate questions, options, and draft wording.
• Use an HR professional or employment law adviser for final decisions in complex or sensitive cases.

A sensible AI policy for SMEs (what to include)

If you want to reduce risk without slowing teams down, a short, practical AI policy should cover:

• What tools are approved and how they’re accessed
• What data must never be entered (e.g., sensitive personal data)
• When human review is mandatory (recruitment, ER, dismissal, redundancy, pay decisions)
• How output is checked and corrected
• A reporting route for misuse (“shadow AI”)

How can PeopleFirstHR help?

PeopleFirstHR supports UK SMEs with selecting and implementing Astute eLearning via VinciWorks, and helping teams roll it out in a way that’s actually adopted (with reporting you can rely on).

Contact us if you would like a demonstration/free trial of VinciWorks Astute e-learning.