AI & Intelligence for Recruitment & Careers — The Practitioner’s Playbook.

SectionThe honest reframe most AI agencies won't tell you

Generic "AI for recruitment" agencies are selling consultancies, search firms and in-house TA teams a CV-screening robot. Drop a CV in, get a yes/no out, save the consultant the reading time. They neglect to mention that fully-automated rejection of a candidate based solely on algorithmic processing is unlawful under Article 22 of the UK GDPR, that an Equality-Act 2010 indirect-discrimination claim turns on disparate impact the vendor has never measured, and that the GLAA, REC and APSCo are all explicit about consultant accountability for hiring decisions.

That is not AI strategy. That is AI cosplay with a regulatory exposure attached.

The high-leverage AI use cases in contingent recruitment, executive search, RPO and in-house talent acquisition are not "auto-reject the bottom 80%." They are: candidate-CV summarisation that saves the consultant 15-20 minutes per shortlist read; role-fit scoring with explainability and a documented human-final-decision step; sales-call summarisation that turns every BD call into a structured client brief; retrieval-augmented generation over your own salary survey and live market data so consultants stop guessing pay benchmarks; ATS-integrated automation across Bullhorn, JobAdder or Vincere so the workflow runs once and updates everywhere; and consultant-reviewed AI-drafted content workflows on the public site.

Each of those is measurable. Each saves consultant hours or lifts placement rate. None of them is "auto-reject candidates." Read the playbook. Run it yourself, or have us ship it on retainer.

SectionThe eight-point audit we run on day one

Score your own AI stack red / amber / green this week. Three or more reds means the foundation is broken — fix that before any new tooling spend.

1. GDPR + Equality-Act-compliant CV summarisation, no automated rejection — Article 22 of the UK GDPR is unambiguous: a candidate has the right not to be subject to a decision based solely on automated processing that produces legal or similarly significant effects. Rejection from a recruitment process is a similarly significant effect. The compliant pattern is summarisation plus surfacing — the AI reads the CV, produces a structured summary, flags fit signals, and the consultant decides. The non-compliant pattern is the AI rejecting candidates and a consultant rubber-stamping the output. If your current stack does the second one, you have a regulatory exposure live today.
2. Role-fit scoring with explainability and human-final-decision — A score on its own is not defensible. Every role-fit score must surface the underlying signals: experience match against the JD, sector adjacency, location, salary band, recency of relevant roles, and any negative signals. The consultant sees the score and the reasoning, then decides. Documented human-final-decision step in the workflow, logged in the ATS. Without that, you cannot defend an Equality-Act 2010 indirect-discrimination challenge.
3. Sales-call summarisation for client briefs — Every BD call and client briefing recorded with consent, transcribed, summarised into a structured brief (role, budget band, must-haves, nice-to-haves, cultural notes, urgency, decision-maker), and written into the CRM as a searchable record. Cuts the gap between client call and a usable brief from "tomorrow" to "ten minutes after the call ends." A 360 consultant can run two more BD calls a week off the time saved.
4. RAG over salary survey and market-data corpus — Retrieval-augmented generation over your own salary survey, live job-board scrapes, market reports, REC pay data and ONS earnings data. The consultant types "what is the right base for a senior data engineer in Manchester with five years of experience?" and gets a sourced, paragraph-cited answer with a confidence score. Not a hallucinated guess from a base model that has never seen your sector or your data.
5. ATS-integrated automation across Bullhorn / JobAdder / Vincere — The AI surfaces are useless if the consultant has to copy and paste between five tools. CV summary writes back to the candidate record. Role-fit score writes to the job record. Call summary writes to the contact record. Outbound sequence triggers from the placement workflow. One workflow, multiple ATS surfaces, version-controlled and tested.
6. Consultant-reviewed AI-drafted content workflow — No unedited AI on the public site for placement-record claims, salary commentary or sector commentary. AI drafts; a senior consultant or director reviews and edits; named-byline publishes. The discipline matters because candidates and clients read these pages closely, and a wrong salary band in a published market report damages the brand the rest of the site is building.
7. Right-to-erasure and data-retention compliance on AI training inputs — Candidate CVs, call transcripts and intake-form data are personal data under UK GDPR. Lawful basis logged. Retention windows defined per data category. Right-to-erasure flow tested end-to-end (the candidate emails, you must remove the data from the ATS, the AI cache, the call-transcription store, and any sub-processor). No candidate data shipped to a model provider that retains training rights. Boring, mandatory, and where most recruitment AI projects quietly fall over the moment a Subject Access Request lands.
8. Productionisation with consultant fallback — Every AI tool has a documented behaviour for when the model is wrong, slow, or unreachable. The role-fit score defaults to "consultant review needed" on low-confidence outputs. The summarisation system flags ambiguous CVs for full read-through. The RAG system surfaces source paragraphs even when the answer is uncertain. Every AI surface has a consultant-fallback runbook tested quarterly. AI without a fallback is a liability waiting for the day the API is rate-limited and the consultant cannot ship the shortlist.

Three or more reds — fix the foundation before commissioning new tooling spend.

SectionSix productised deliverables we ship per cycle

On a Foundation, Compound or Architect retainer, the same six outputs land in your portal each cycle. Industry-tuned, fixed scope, dated, owned by you.

GDPR-compliant CV summarisation. A consultant uploads or syncs a candidate CV from the ATS. The pipeline returns a structured summary: name, current role, total experience, sector experience, skills extracted, location, salary expectation if disclosed, a fit-signal ribbon against the live brief, and any negative signals (gaps, frequent moves, sector mismatch). Article 22 of the UK GDPR compliant by design — the AI summarises and surfaces, the consultant decides. Right-to-erasure tested end to end before go-live. Time to first signal: 14-21 days. Owned by you, exported as written SOP plus the orchestration scripts.

Role-fit scoring with explainability. Every candidate against every live role gets a fit score from 1-5 with an explanation: experience match against JD, sector adjacency, location fit, salary band overlap, recency of relevant work, and explicit negative signals. The score and the reasoning surface together in the consultant's ATS view. Logged audit trail of the score, the reasoning, and the consultant's final decision — defensible against an Equality-Act challenge. Built on public LLM providers with documented prompts and version control. Time to first signal: 21-30 days.

Sales-call summarisation for client briefs. Every recorded BD call or client briefing transcribed with consent, summarised into a structured brief (role, budget, must-haves, nice-to-haves, cultural notes, urgency, decision-maker, next step), and written automatically into the CRM as a searchable record. Consultant reviews, edits if needed, and ships the brief to the delivery team within thirty minutes of the call rather than the next morning. Drives a measurable lift in BD throughput by recovering the post-call write-up hour.

RAG over salary survey and market-data corpus. A retrieval-augmented generation system trained on your salary survey, sector market reports, REC pay data, ONS earnings data and your own placement history. Consultants ask pay-benchmarking and market questions and get a paragraph-cited answer with confidence score and source documents. Public RAG patterns over public LLM providers. Cuts misquoted pay benchmarks at briefing stage to near-zero, which is where most reputational damage on the consultant relationship happens.

ATS-integrated automation. Whichever ATS you run — Bullhorn, JobAdder, Vincere, RecruiterFlow, JobScience, Adapt — we wire the AI surfaces directly into the candidate, job, contact and placement records. CV summary writes to the candidate. Role-fit score writes to the candidate-job match. Call summary writes to the contact. Outbound sequence triggers from placement workflow. One workflow, version-controlled, tested end-to-end with rollback documented. The difference between "we have AI" and "AI is part of the consultant's daily flow."

Consultant-reviewed AI content workflow. A documented draft-to-publish workflow: AI drafts a sector commentary, salary trend piece or candidate-attraction guide; a senior consultant or director reviews and edits; named-byline publishes. Author schema points at the consultant. Throughput rises 3-5x versus all-human drafting; technical and commercial accuracy stays at the level a candidate or client would read with confidence. Never unedited AI on placement claims, salary data or sector commentary.

SectionWhat to do this week

Three actions, ranked by leverage. Same first three steps we ship in week one of a Foundation retainer for a recruitment business.

1. Time the lag between a BD call and a usable client brief. Owner: founder or head of BD. Time: one afternoon. Pull yesterday's BD calls. How long between call ending and a structured, searchable client brief existing that a delivery consultant could action? If it is more than thirty minutes — or, in most consultancies we audit, the next morning at best — sales-call summarisation is your highest-leverage AI deployment.
2. Audit your CV-screening flow against Article 22 of the UK GDPR. Owner: managing director or compliance lead. Time: 30 minutes. Walk through what happens to a CV after submission. Does any candidate get rejected without a named consultant making the final decision and logging it? If yes, you have a live regulatory exposure under Article 22 — fully-automated rejection of candidates is unlawful. The fix is not "less AI" — it is making the consultant's review and decision the documented, logged final step.
3. Decide DIY, DWY or DFY for the next 90 days. Owner: founder. Time: 30-min discovery call. We will confirm the right way in writing within two business days. See the three ways.

SectionFive questions recruitment operators ask us about AI

Article 22 of the UK GDPR — does it really stop us auto-rejecting candidates? Yes, in plain terms. Article 22 gives every candidate the right not to be subject to a decision based solely on automated processing that produces legal or similarly significant effects on them, and rejection from a recruitment process is a similarly significant effect. The compliant pattern is AI summarisation and surfacing, with a named consultant making and logging the final decision. The non-compliant pattern is "the model rejected this candidate" with no meaningful human review. The ICO has been explicit, and an REC member firm running auto-rejection at scale is one Subject Access Request away from a defensible regulatory file. This is not theoretical risk — it is a live exposure in current case law.

Is role-fit scoring really defensible if a candidate challenges us under the Equality Act? Defensible if you can show three things: the score has explainability — the consultant sees the underlying signals and the reasoning, not just a number; the consultant makes the final decision and that decision is logged with their name against it; and you have monitored the score for disparate impact across protected characteristics and corrected where needed. Score-only-no-reasoning, with no human-final-decision and no disparate-impact monitoring, is indefensible. Score-plus-reasoning-plus-human-decision-plus-monitoring is defensible and is the pattern we ship.

Bullhorn versus JobAdder versus Vincere — does the ATS choice change what AI we can run? The ATS choice changes the integration plumbing, not the AI surfaces themselves. Bullhorn has the most mature API and the largest marketplace. JobAdder is leaner, fast to wire up, popular with mid-market UK and AU agencies. Vincere has strong native automation and a healthy webhook story. RecruiterFlow, JobScience and Adapt are all serviceable. The same six deliverables — CV summarisation, role-fit scoring, call summarisation, RAG, ATS-integrated automation and consultant-reviewed content — run on any of them. We have shipped on Bullhorn and JobAdder most often; the timeline is similar across the major three.

Is consultant review on AI-drafted content really necessary, or is that a CYA exercise? It is necessary for two distinct reasons. The first is commercial accuracy: candidates and clients read your salary commentary, sector reports and market guides closely, and one wrong pay band in a published piece undermines the consultant relationship for the next eighteen months. The second is brand and regulatory: REC and APSCo are explicit about member accountability for published claims. Consultant review on AI drafts is not a CYA box-tick — it is the production gate that lets you ship 3-5x more content without ever shipping a number a candidate or client could fact-check and challenge.

Can we run this ourselves with the playbook plus £750 audit? Yes. Most of the audit-and-fix list above is achievable in-house if you have an operations lead plus a developer or systems integrator half-week per cycle plus a senior consultant who can spend an hour a week reviewing AI drafts. The £750 audit gets you a written red / amber / green of all eight points, a prioritised next-step list with named owners and dates, and a copy of the workflow templates and prompt patterns we use ourselves. If you sign for DWY or DFY within 30 days, the audit fee credits against the first cycle.

SectionWhere to go from here

If you want this shipped end-to-end on a productised retainer, book a 30-minute discovery call. Tailored proposal in writing within two business days.

If you would rather have a senior practitioner reviewing your team's AI deployments each week, the coaching plans start at £750/month with rolling cycles and walk-away rights. If you have a hard deadline — an ATS migration that needs the AI surfaces wired in before go-live, or a new-vertical AI pilot you need running before a peak hiring window — the two-week embedded sprint lands a senior practitioner inside your tools for ten working days at £3,000 fixed.

Or run it yourself. Read this playbook end to end, run the eight-point audit, ship one deliverable a month for six months. Twice-quarterly office hours are open to anyone using the playbooks — bring your work, get reviewed, no charge.