Insights · Technology

What we automate — and what stays human

Recruitment has two kinds of work: administration that slows people down, and judgement that cannot be delegated. We automate the first. The second stays with people.

There is a version of “AI in recruitment” that should worry everyone: software screening people out of consideration before a human being has read a word. Candidates eliminated by keyword matching, ranked by opaque scoring, rejected at 3am by a system nobody reviewed. That is not what we do.

What the technology does

Our workflow uses AI and automation for the parts of recruitment that were always clerical work dressed up as judgement:

  • Capturing experience in the candidate’s own words. Guided digital capture replaces the recruiter paraphrasing a phone call and typing it up with errors — no intermediary, no transcription, no distortion.
  • Briefing candidates properly. Requirement documents are summarised into clear, accurate briefs, so applicants understand precisely what is being asked before they invest their time — and ours.
  • Keeping people informed. Templated-but-personalised updates at each stage — each one reviewed against the individual before it is sent.
  • Checking our own paperwork. Our documents are checked for consistency and completeness before they leave us — so an assessor’s questions are about the candidate, not about errors.

One rule sits above all of it: we only use commercial, paid AI services whose terms guarantee that nothing we process is used to train their models. What you share with us stays with us. And the division of labour is strict — automation prepares; people decide. Every application is read by a practitioner, and every shortlist is built by one.

What stays human

Technology does a great deal here. It does not choose people.

Selection is judgement: weighing evidence against what a team actually needs, knowing when a non-obvious candidate is the right one. That is the part of recruitment that takes years to learn, and it is the part we keep with practitioners — every application read by a person, every selection decision made by one.

We will not pretend to know what good practice will look like in five years. The technology is moving quickly, and we can imagine a time when parts of selection, properly governed and tested, are better done with AI assistance than without. If our practice ever changes, we will say so — openly, here — and any such step would be governed, sampled and accountable, never introduced silently. What does not change is the discipline behind every decision: governed tools, tested outcomes, and honesty about which is which.

The result is not a slower process. It is a faster one with the judgement left in: administration moves at technology speed, decisions move at practitioner speed, and candidates always know which is which.