CV and application parsing
Turning free-text CVs and application forms into structured fields (skills, tenure, qualifications) so every applicant is compared on the same criteria rather than on layout or formatting.
— AI use case —
AI for recruitment
screening.
AI can make recruitment screening faster and more consistent, but it puts an algorithm between a person and a job. The Institute of AI helps UK organisations get the benefit without inheriting the bias, the legal exposure, or the reputational risk. Independent, and with no recruitment software to sell.
— State of play —
A real use case,
handled with care.
Recruitment screening is one of the first places most UK organisations meet AI in a decision that materially affects a person. High application volumes, structured CVs, and clear pass or fail moments make it an obvious candidate for automation, and tooling now reaches from job boards through applicant tracking systems into interview scheduling.
The upside is real. Screening that once took recruiters days can be triaged in minutes, freeing skilled people to spend time with the strongest candidates rather than the pile. Done well, structured AI screening can also be more consistent than a tired human reading the four hundredth CV of the week.
The risk is equally real, and it is not theoretical. A screening model learns from your past hiring, so it can quietly reproduce the very patterns you are trying to move away from. Because a rejected applicant rarely finds out why, unfair screening is unusually hard to detect from the outside, which is exactly why it needs governing from the inside.
Nine
protected characteristics under the Equality Act 2010, the legal frame every screening model sits inside
Four
levels of professional accreditation, Student to Fellow, against IoAI's published competency framework
Free
to sign the UK AI Readiness Charter, free for any employer
— Use cases —
Where AI helps with
recruitment screening.
The strongest screening use cases order and speed up human review rather than replace it, and keep a written, auditable basis for every score.
Ranking and shortlisting
Scoring applicants against the requirements of a specific role to surface a longlist. Best used to order a human review, not to auto-reject, so a person still owns the cut line.
Knock-out criteria checks
Confirming genuine, role-essential requirements such as a mandatory professional certification, where the rule is objective and documented rather than inferred from correlations. Right-to-work stays on the prescribed Home Office routes, not an AI check.
Skills and assessment scoring
Marking structured exercises, coding tasks, or situational judgement tests consistently at scale, with the scoring rubric written down and auditable rather than hidden inside the model.
Job description review
Checking adverts and role specs for exclusionary or gendered language before a vacancy goes live, widening the pool at the top of the funnel instead of filtering it at the bottom.
High-volume triage and routing
Sorting seasonal or graduate intakes, deduplicating applicants, and routing enquiries so recruiters reach real candidates faster, with clear signposting that AI is part of the process.
— What must be governed —
Faster hiring,
not blind hiring.
Screening decisions affect people and carry legal duties. These are the questions to settle before a model touches a live vacancy, tailored to recruitment.
01
Model risk and validation
Prove it works before it screens anyone
A screening model is only as sound as the data it learned from. Validate it against real outcomes, test how it behaves on edge cases and sparse CVs, and document what it is and is not fit to decide. A model that has never been validated should never sit in front of a live vacancy.
100%
of screening models validated before go-live
02
Fairness and bias
Test for disparate impact by protected group
The Equality Act 2010 makes indirect discrimination unlawful even when it is unintended, and bias monitoring is the employer's duty. Test selection rates across the characteristics you lawfully hold monitoring data for, and test proxies such as postcode, career gaps, or the name of a school regardless, since they can stand in for characteristics you cannot measure directly.
Tracked
selection rates across the groups you lawfully hold data for
03
Explainability
Be able to say why someone was screened out
If a candidate, a tribunal, or your own board asks why an applicant was rejected, ‘the algorithm decided’ is not an answer. Screening decisions need reasons a person can understand and defend, which rules out opaque scoring where the drivers cannot be surfaced.
0
unexplainable rejections in the pipeline
04
Data protection
UK GDPR applies from the first CV
Screening is large-scale processing of personal data, so it needs a lawful basis, a Data Protection Impact Assessment, and clear retention limits. UK GDPR gives candidates safeguards around solely automated decisions with significant effects, including a route to human review; the precise duties turn on current data protection law, so settle them before the first CV is processed.
1
DPIA completed before processing begins
05
Human accountability
A named person owns the decision
Accountability for a hiring outcome cannot be handed to a vendor or a score. Keep a meaningful human in the loop at the point of rejection, and name the person answerable for the process so responsibility is settled long before anyone challenges an outcome.
1
accountable owner per screening process
06
Monitoring after go-live
Watch for drift once it is running
A model that was fair at launch can drift as roles, applicant pools, and labour markets change. Set a monitoring cadence, track selection rates over time, and define the trigger that pauses the system. Governance is a standing commitment, not a launch-day certificate.
90
days is a sensible maximum between fairness reviews
— How the Institute of AI helps —
One independent body,
three ways to help.
Advice, engineering, and professional standards from a single independent body. IoAI has no recruitment software to sell, so the guidance is yours, not a sales route.
Consultancy
Independent advice
Standards-led advisory for HR leaders and boards deciding whether and how to screen with AI. We help you assess a tool, run the DPIA and fairness testing, and set governance that stands up to scrutiny, with no recruitment software to sell.
- Vendor and tool assessment
- DPIA and bias testing support
- Screening governance that holds up
Solution design
Systems built and handed over
The Institute of AI's engineering practice designs and builds screening support around your controls, with explainability and audit logging in from the start, then hands the system to your team with the knowledge to run it.
- Explainable, auditable by design
- Built around your controls
- Knowledge transfer to your staff
Professional body
Standards and the Charter
As the UK's professional body for AI, IoAI accredits the AI professionals who build and govern your screening, and maintains the UK AI Readiness Charter so your commitments to fair, accountable hiring are visible and free to sign.
- Professional accreditation and CPD
- The UK AI Readiness Charter
- Organisation-level recognition
Screen with AI,
and stay answerable.
A short call with the Institute of AI. Plain answers on where AI fits your recruitment screening, what to govern, and a clear next step. No recruitment software to sell.
— Go deeper
— Sectors
Sector programmes
How the Institute of AI supports your industry, grounded in the UK regulation that applies to it.
— Consultancy
AI consultancy
Independent, vendor-free advice for leaders putting AI to work in decisions that affect people.
— Charter
The Charter
The UK AI Readiness Charter: five pledges for getting your organisation ready for AI, of which you sign at least three. Free for any organisation.