USE CASE · RECRUITMENT
More time with candidates and clients. Less time in the ATS.
The AI Recruitment Worker parses and summarises applications against the spec, re-discovers candidates already in your database, drafts candidate messages and keeps your ATS up to date. Your recruiters make every hiring decision.
PROBLEM
Recruitment is a relationship business, but consultants spend much of their day on admin.
- One advert brings in hundreds of CVs, most of them read in seconds and the good ones easily missed.
- Your database holds thousands of candidates who never get found again.
- ATS notes are patchy, so handovers and client updates take longer than they should.
- Interview scheduling turns into a long email chain.
- Candidates wait for replies, and your reputation takes the hit.
Generic AI tools don't help here. They don't connect to Bullhorn or Vincere, and pasting CVs into a chatbot creates a data protection problem.
CURRENT MANUAL WORKFLOW
- 1.Applications arrive from job boards and email.
- 2.A consultant opens each CV and skims it against the spec.
- 3.Promising candidates get a call, and notes go in a notebook or a half-filled ATS field.
- 4.The ATS is searched by keyword, if anyone has time.
- 5.Updates and rejections go out late, or not at all.
- 6.Interviews are arranged by email between candidate, client and consultant.
- 7.Client summaries are written from scratch for each candidate.
AI-OPERATED WORKFLOW
Intake
The worker picks up every application from the ATS, job boards or the inbox.
Parse
It creates a clean, de-duplicated candidate record in your ATS.
Summarise against the spec
It produces a short summary of matches, gaps and questions to ask, quoting evidence from the CV.
Rank for review
It orders candidates for each role so the recruiter starts with the strongest. It never rejects anyone itself.
Re-discover
It searches your existing database for past candidates who fit, with reasons.
Draft messages
It writes acknowledgements, invites and updates in your tone for recruiter approval.
Coordinate interviews
It proposes slots across calendars and drafts confirmations and reminders.
Log
It turns screening-call notes into structured ATS updates and client-ready summaries.
HUMAN CONTROL
The worker prepares. Recruiters decide.
Approvals
Every candidate- and client-facing message, and every status change, is approved by a recruiter. No solely automated hiring decisions.
Review queue
A ranked list per role, each candidate with an evidence-based summary, to accept, reorder or reject.
Audit trail
Every summary, ranking, message and ATS write is logged with its inputs, criteria, approver and timestamp. This supports candidate data requests and fairness reviews.
Fairness controls
Only the job-relevant criteria you set are used. Protected characteristics and obvious proxies are excluded, and rankings are reviewed for bias.
Escalation
Edge cases, data-subject requests and anything outside policy go to a named person.
IMPLEMENTATION APPROACH
We start with one workflow, usually CV summaries and ranking into the ATS, and get it into production first.
| Phase | What happens | Timeline |
|---|---|---|
| 1. Workflow review | You show us your process, ATS and a few live roles. We agree scope, success measures and data access. | Week 0 (one call plus a short written scope) |
| 2. Discover & design | We baseline screening time and response times, define criteria and fairness rules, and build an evaluation set from past roles and outcomes. | Week 1–2 |
| 3. Build & integrate | We connect the ATS, inbox and calendars, then build the worker, review queue and logging. | Week 2–4 |
| 4. Shadow mode | The worker summarises and ranks live roles alongside your consultants. Nothing is sent and no statuses change. We compare against their judgement. | Week 4–5 |
| 5. Go live & operate | We go live role by role or desk by desk, with monitoring and periodic bias review. Then we extend to re-discovery, scheduling and more. | Week 6 onwards |
Phases 2–5 are delivered as a Production AI Sprint. Production AI Sprint
FAQ
- Does the AI reject candidates?
- No. It summarises and ranks, and a recruiter makes every shortlist and rejection decision. That's how we design it under UK GDPR rules on automated decision-making, and it's how good recruiters want to work anyway.
- Which ATS does it work with?
- Agency systems like Bullhorn, Vincere, JobAdder, Firefish and Mercury, and in-house platforms like Greenhouse, Workable, Teamtailor and Ashby, through their APIs. We confirm access for your setup in the workflow review.
- How do you handle candidate data and UK GDPR?
- Candidate data stays in your ATS and a deployment you control, in your cloud tenancy or a UK/EU-region environment, using model APIs that don't train on your data. Every processing step is logged, and we support your DPIA.
- What about bias?
- Screening uses only job-relevant criteria you define. Protected characteristics and obvious proxies are excluded, and rankings are reviewed against recruiter decisions in shadow mode and periodically after go-live.
- Will it message candidates on its own?
- Not by default. Messages are drafted for recruiter approval. Some agencies later allow simple acknowledgements to send automatically. That's your choice, and it's logged.
Show us how a role moves through your desk.
Take one live role from advert to placement and walk us through it. We'll tell you plainly where an AI worker would save consultant time, what must stay with recruiters, and what the first production version would look like.