When should an SME use generic AI for HR and when should it stop?
The question is no longer whether employees and managers will use AI. They already do. The real management challenge is deciding which tasks are suitable, what information may be shared and where human judgement is non-negotiable.
AI access is not the same as HR capability
Generic AI is excellent at language: summarising, restructuring, explaining and drafting. That can make a manager faster. But an articulate answer can conceal a missing fact, an outdated assumption or a failure to recognise legal risk. HR decisions are particularly sensitive because context, consistency and the effect on a real person matter.
A useful policy therefore governs decisions, not just tools. It should state the approved uses, prohibited data, required review and escalation triggers.
A traffic-light model managers can remember
GREEN - use AI for agendas, neutral rewrites, anonymised checklists, summaries of your own non-sensitive notes and brainstorming questions.
AMBER - use only as a reviewed draft for letters, policies, performance plans, redundancy materials or interpretations of legal process. Verify sources and dates.
RED - do not delegate credibility findings, dismissal decisions, legal conclusions, medical judgements or decisions involving discrimination, whistleblowing, safeguarding, TUPE or collective redundancy.
Never enter more personal information than the task requires. Names, health information, allegations, union membership and other sensitive details need especially careful handling. Anonymisation is useful only if a person cannot reasonably be re-identified from the remaining context.
Why specialist design matters
A general chatbot is optimised to be helpful across almost any topic. A responsible specialist system should also know when helpfulness means slowing down, asking a missing question, exposing uncertainty or recommending human support.
Phronesis is built around that principle. It brings UK HR guardrails, ACAS-aligned process prompts, source citations, confidence information, currentness indicators and red-flag escalation into the conversation. It does not make the final employment decision.
What changes for SMEs over the next few years
AI will lower the cost of basic knowledge and first-draft administration. The premium will shift toward judgement, evidence, implementation and human trust. Outsourced HR models based mainly on answering routine telephone or email questions will face pressure; advisers who combine technology with expert intervention will become more valuable.
For SMEs, this is an opportunity to build management capability earlier. A 20-person firm can give managers access to consistent guidance that once required a larger HR function. The winners will not remove people from people management; they will reduce waiting, improve preparation and spend human expertise on the moments where it matters.
Adopt AI with guardrails, not guesswork
Use the free Phronesis plan for a real, anonymised everyday HR question, then compare the experience with a generic chatbot. Explore how it identifies risk and next steps at https://www.phronesis.org.uk/how-phronesis-answers.
Useful sources
https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/
https://www.gov.uk/government/publications/ai-adoption-research/ai-adoption-research
Important: This article provides general HR information, not legal advice. Check current official guidance and obtain professional advice for high-risk or fact-sensitive decisions.