SCOPE NOTE
This post provides general context on UK data protection, automated decision-making and responsible AI in recruitment. It does not constitute legal advice. For guidance specific to your organisation, consult a qualified employment law or data protection specialist.The automated decision-making provisions introduced through section 80 of the Data (Use and Access) Act 2025 created Articles 22A–22D of the UK GDPR framework. The relevant provisions came into force on 5 February 2026.
AI recruitment is most useful when it removes repetitive work, improves the quality of evidence available to hiring teams and leaves important judgement with people. For a 51–200-person UK business, the goal should not be to automate every stage of recruitment. It should be to decide which tasks AI can support, which decisions require human review and how the organisation will measure whether the technology is actually improving hiring.
AI recruitment uses artificial intelligence to support tasks across sourcing, screening, assessment, interviewing, candidate communication and recruitment administration.
That can include generating a job advert, identifying potential candidates, extracting skills from CVs, suggesting interview questions, summarising interview evidence or answering routine candidate questions. The UK Government’s Responsible AI in Recruitment guidance identifies sourcing, screening, interview and selection as areas where organisations are already using AI-enabled technologies.
The important distinction is between a task and a decision. Drafting a job advert is a task. Deciding whether someone should progress to a final interview is a significant hiring decision. The closer an AI system gets to determining a candidate’s progression, the more important human oversight, transparency, testing and safeguards become.
For an SME, this distinction is practical. AI can help your People team process information. It should not become an unexplained gatekeeper between qualified candidates and your hiring manager.
Start with high-volume, repeatable tasks where the consequences of an error are relatively easy to detect and correct.
A useful way to assess an AI recruitment tool is to ask four questions: What task is being automated? What evidence does the system use? What happens if it is wrong? Who reviews the output?
| Recruitment stage | Appropriate AI role | Human responsibility |
| Job description | Drafting, structure and language checks | Define outcomes, skills and genuine requirements |
| Sourcing | Finding potential matches | Check relevance and broaden the search where needed |
| CV screening | Extracting evidence and organising applications | Review criteria and borderline candidates |
| Interview design | Generating structured questions | Confirm questions test the required competencies |
| Interview assessment | Structuring notes and evidence | Interpret context and candidate responses |
| Shortlisting | Prioritising information for review | Decide who progresses |
| Final selection | Comparing documented evidence | Make the hiring decision |
| Candidate communication | Drafting routine messages | Handle sensitive or consequential conversations |
This is where many organisations make the wrong first move. They buy a tool because it promises faster screening, before defining what a good candidate looks like.
The 2026 CIPD report found that 46% of organisations still take an ad hoc approach to recruitment, while only 19% calculate the return on investment of their recruitment processes. If the underlying process is inconsistent, automating it can make the inconsistency faster rather than fix it.
Before buying AI, define the role, success measures, essential skills, desirable skills, assessment criteria and decision owners. Then decide which parts of that process benefit from automation.
Human judgement should remain strongest where context, fairness, candidate circumstances or the final employment decision matter.
An AI system can identify patterns in application data. It cannot automatically understand every relevant piece of human context.
Consider a candidate whose CV uses different terminology from the job description. A keyword-driven system may miss relevant experience. Consider another candidate who has taken a career break, changed sector or gained valuable skills outside a conventional career path. A rigid model may treat an unusual career pattern as a weakness when a human reviewer would ask a better question.
The UK Government’s responsible AI guidance warns that recruitment technologies can create or amplify bias, digital exclusion and discriminatory advertising or targeting. It recommends organisations assess risks, test systems, maintain appropriate governance and provide meaningful human oversight.
The ICO’s Recruitment Rewired work adds another practical warning. Its findings, based on evidence from more than 30 employers that engaged with the regulator between March 2025 and January 2026, focus on meaningful human involvement, transparency and safeguards in automated recruitment decision-making.
Human oversight therefore needs to be real, not ceremonial. A manager who simply accepts an AI-generated shortlist without understanding how it was produced is not providing meaningful review.
AI can help teams handle application volume, but it cannot compensate for a poorly defined role or weak selection criteria.
The latest CIPD data illustrates the problem. While 58% of employers reported an increase in unsuitable applicants, only 33% reported an increase in suitable applicants. For a small People team, that creates pressure to process more information without necessarily improving the quality of the shortlist.
AI can help organise that information. But there is a danger in treating application volume as the problem.
If the job description is too broad, AI can process a larger pool of irrelevant applications. If the essential criteria are poorly chosen, an automated screen can consistently prioritise the wrong signals. If the assessment measures generic performance rather than job-relevant capability, faster scoring simply produces faster answers to the wrong question.
This is why I would start with the hiring process rather than the technology.
Ask: What evidence would convince us that this person can succeed in this role?
Then build the recruitment process around that evidence.
There is another 2026 complication. Candidates are using AI too. The CIPD found that 27% of organisations that attempted to recruit reported excessive use of generative AI in applications. Of the employers that attempted to monitor candidate AI use, 75% said they had rejected candidates for their use of AI.
That creates a two-sided AI recruitment system. Employers are using AI to assess applications while candidates are using AI to create them. The answer is not necessarily to ban AI. It is to strengthen the parts of selection that test genuine capability: structured interviews, relevant work samples, job-specific assessments and evidence-based evaluation.
Treat AI recruitment as a process change first and a technology purchase second.
1. Define the problem. Do you have too many applications, slow scheduling, inconsistent screening, weak candidate communication or poor evidence at interview? Pick one problem.
2. Choose the lowest-risk task first. Routine administration, drafting and information organisation are usually easier places to pilot than final candidate selection.
3. Define the human checkpoint. Write down who reviews the output, what they are expected to check and what happens when the AI output looks wrong.
4. Pilot before scaling. The Government’s responsible AI guidance recommends testing systems in the organisation’s real-world environment before deployment at scale. Test against different candidate profiles and look for errors, accessibility problems and inconsistent outcomes.
5. Tell candidates where AI is being used. Transparency matters. Candidates need to understand when an AI-enabled system is part of the recruitment process, particularly where it influences their progression.
6. Measure the result. Do not stop at time saved. Track whether the quality of shortlists, candidate experience, diversity, interview conversion and eventual hiring outcomes improve.
The ICO also recommends organisations consider how applicants can raise concerns and how human review can be provided where decisions or recommendations are questioned.
For a smaller organisation, this does not require a huge AI governance department. It requires clear ownership, documented decisions and a willingness to stop using a tool when the evidence shows that it is not performing as expected.
Measure the quality of decisions and outcomes, not simply the number of tasks automated.
Start with a baseline before introducing the tool.
Useful measures include:
The 2026 CIPD research is a useful reminder that measurement is still a weakness. Only 19% of organisations calculate the ROI of recruitment processes.
That means an SME can create a competitive advantage simply by being more disciplined about evidence.
If an AI screening system saves six hours a week but removes strong candidates from consideration, it is not delivering the result the business needs. If it reduces administration while preserving candidate quality and improving the hiring team’s capacity for better conversations, the picture is different.
The question should always be: did the recruitment outcome improve?
Ask the supplier for evidence about accuracy, limitations, testing, data use, human oversight and how the system behaves in your organisation.
The UK Government recommends asking suppliers for evidence supporting claims about performance, ROI, efficiency and fairness. Depending on the system, that may include impact assessments, risk assessments, model documentation and a data protection impact assessment.
For an SME, I would put these questions in front of every supplier:
If the answers are vague, that is useful information in itself.
AI recruitment should make a hiring process more evidence-based, not less accountable.
The best next step for a UK SME is therefore not to ask, “Where can we add AI?” Start with the hiring decisions that matter, identify the repetitive work surrounding them, and then decide where technology can genuinely improve the process.
AI can help your team move faster. Your hiring system still needs to know where it is going.
AI recruitment uses artificial intelligence to support tasks across sourcing, screening, assessment, interviewing, candidate communication and recruitment administration. The appropriate level of automation depends on the task, the data involved and the consequences of an error.
Common uses include drafting job descriptions, finding potential candidates, organising CV information, generating structured interview questions, summarising evidence and handling routine communication. Human review remains important for consequential hiring decisions.
AI can reduce repetitive administration, help teams handle application volume and create more consistent recruitment workflows. Its value should be measured through outcomes such as shortlist quality, candidate experience, hiring speed and quality of hire rather than automation volume alone.
Risks include biased or incomplete data, discriminatory outcomes, digital exclusion, weak transparency and over-reliance on automated recommendations. UK Government guidance recommends testing, governance, transparency and appropriate human oversight.
AI can automate or support parts of recruitment, but it does not remove the need for human judgement. People still need to define the role, assess context, review evidence, make consequential decisions and manage candidate relationships.
Start with a clearly defined problem, choose an appropriate low-risk task, establish a named human owner, pilot the technology, tell candidates where AI is being used and monitor outcomes. For regulatory questions specific to an organisation, obtain qualified professional advice.
There is no single rule that makes AI recruitment simply legal or illegal. Organisations need to consider data protection, equality, transparency, automated decision-making and other relevant obligations according to how the system is used. The ICO and UK Government have published guidance, and organisations should obtain specialist advice for their circumstances.
Sabiha is a Talent Acquisition Director, Speaker, and Author with 16+ years of experience helping UK organisations build smarter, more inclusive hiring systems. She is the author of How to Use AI to Win Talent and Retain People (Trotman, 2026), a CIPD-aligned guide for HR leaders navigating AI in the workplace. Connect with her at meetsabiha.com.
Global Talent. Ethical AI. Strategic Hiring. Sustainable Retention.
