AI for Recruitment Agencies: The Wins, and the One Area the Law Watches

Recruitment is a language business. Job ads, CVs, outreach messages, interview notes, candidate updates — nearly everything an agency produces or processes is text. That makes it one of the industries where today's AI lands hardest, and one of the few where the law has something specific to say about it. Both halves matter.

The work AI does well in an agency

Job ads and outreach that don't sound like everyone else's. Give an assistant your client's brief, your tone guidelines, and three of your best-performing past ads, and it will draft a strong first version in seconds. Same for candidate outreach: personalised messages drafted from the CV and the role spec, reviewed and sent by a human. Consultants who batch this recover hours a week.

Interview and call notes. A transcribed intake call becomes a structured brief — must-haves, nice-to-haves, salary range, red flags — filed against the role in your CRM without anyone retyping it. Candidate interviews become searchable summaries. This is the least contested win in the industry: nobody misses writing up call notes.

CRM hygiene. Every agency CRM is a graveyard of half-filled records. An assistant connected to your CRM can chase the gaps — pulling details from email threads and CVs into the fields where they belong, flagging duplicates, updating statuses from correspondence. Connected properly, via the kind of integration in our MCP server directory, it acts on the CRM directly rather than producing text for someone to paste.

Candidate communication. The updates that don't happen — "you're still in the running", "the client pushed the decision to next week" — are the biggest driver of candidate drop-off. Drafted automatically from pipeline changes and approved in batches, they actually get sent.

The part the law cares about: screening

Here is where recruitment differs from most industries. Under the EU AI Act, AI systems used for recruitment and selection — screening applications, evaluating candidates — are classed as high-risk. If your agency places candidates with EU clients, or screens for roles based in the EU, the Act can reach you even from the UK. Those high-risk duties were pushed back by the Digital Omnibus and now bite from 2 December 2027 — which is breathing room, not a reprieve, because the obligations (human oversight, logging, telling people the system is in use) describe how a defensible screening process should already work. UK-only agencies aren't off the hook either: UK GDPR already governs automated decision-making about individuals today, and the ICO has taken a close interest in AI in hiring.

The practical line to draw: use AI to prepare screening decisions — summarising CVs against the brief, surfacing evidence, flagging gaps — and keep a human making the actual call, visibly and on the record. "The model rejected them" is a sentence you never want to say to a regulator, a tribunal, or a client. Automated shortlisting with no meaningful human review is exactly the deployment that attracts scrutiny, and the accuracy isn't there for it anyway: models inherit the biases in their training data, and hiring is where those biases carry legal consequences.

Handled that way, AI in screening is defensible and genuinely useful — it reads the hundredth CV with the same attention as the first, which is more than any human screener can claim by Friday afternoon.

What it costs

The general SME cost picture applies to agencies without much adjustment. Off-the-shelf assistants at £15–£60 per consultant per month pay for themselves on job ads and notes alone. The bigger step — AI connected to your actual CRM and inbox with permissions and an audit trail — is a project, typically £5,000–£25,000 depending on your stack, and it's where the CRM-hygiene and candidate-comms wins live. Most recruitment stacks (Bullhorn, Vincere, JobAdder, or a general CRM like HubSpot) can be connected; the pattern is described in how to connect AI to your business tools without writing code.

Where to start

Not with screening — with notes. Call and interview write-ups are high-volume, rules-based, and nothing breaks if a summary needs editing: the ideal first process by the criteria we use everywhere. Prove the value there, then widen into CRM hygiene, then candidate comms, and treat anything touching selection decisions as a governance project as much as a technical one.

If you want the full picture first — which processes in your agency would pay back fastest, and in what order — that is what AI Mapping produces. Or start with the two-minute readiness assessment.