78 Rejections. The Pattern Nobody But Our AI Agent Spotted.

A thought experiment, not a case study. This is where we think agent-based AI is heading for staffing, told as a story, because the point lands harder that way.

78 Rejections. The Pattern Nobody But Our AI Agent Spotted.

Introducing Dominic

Dominic had been on the database for 43 weeks. A systems architect, solid track record, nothing wrong with him on paper.

In that time, he made 78 shortlists. He got zero offers.

Not because he wasn't good enough. Because every hiring manager who saw his CV saw a generalist; competent, unremarkable, easy to rank third and forget. Nobody had ever framed him as anything else.

This is the bit that should bother every consultant reading this: you've got a Dominic on your books right now. Somebody who keeps losing, not on merit, but on presentation. And you probably don't know it, because nothing in your system is built to notice a pattern like that. It's not one bad shortlist. It's 78 of them, spread across months, invisible unless something is watching for exactly that shape of failure.

That's the job we think AI agents are actually suited to. Not replacing judgement. Spotting the pattern a person would never have time to look for.

Introducing Marika - a Kyloe AI agent of the future

Picture an agent (we’ll call her Marika) sitting across a recruitment database, doing the unglamorous work: batch queries, placement histories, shortlist outcomes. Nothing exciting, until she asks the question that a busy consultant never gets round to: why do some candidates always come second.

She finds Dominic. And 46 others like him. Strong technical matches, consistently overlooked, no fault on their record.

Here's the part that matters: she doesn't quietly go and fix it herself. She flags it. To the team, in plain language, with the evidence attached. Then she proposes something specific. Not “let me sort this out,” but a structured plan, three specialist agents, each doing one job well:

  • One handling temp and contract roles, where speed decides everything.
  • One working permanent placements, where a longer, more considered conversation is what's missing.
  • One doing nothing but reframing - taking a candidate's real experience and presenting it the way it should have been presented the first time.

Dominic goes from “generalist, rejected three times” to “legacy platform transformation specialist.” Same person. Different frame. Two hiring managers call within 48 hours.

The bit that matters most

None of this replaces the consultants. It runs alongside them.

The agents surface the pattern and do the volume work. The humans do what humans are actually good at; reading a hiring manager, having the awkward conversation, knowing when a candidate needs pushing and when they need protecting. Marika doesn't have that. She was never going to. That's not the job she's built for.

What you get when you put the two together isn't AI recruitment. It's recruitment that finally has time to look at the candidates it was quietly failing.

A happy ending

Over 30 days, that team places 89 people. Churn comes in at 4%, against an industry average of 22% for that cohort. Not because the AI is clever. Because it did the one thing nobody had time to do - look for a pattern across hundreds of outcomes and then hand the result to people who knew what to do with it.

Whatever the surplus from that kind of efficiency gets spent on next (better tools, more time, something that gives back or contributes to society) is a conversation worth having. But it's a separate conversation. The one that matters today is simpler.

The moral of the story

Good candidates get missed because they're framed wrong, not because they're weak. That's exactly the kind of pattern AI is suited to catching, working alongside your consultants, not instead of them.

But it’s important to look at a second point here. Should you choose one AI that does everything to an OK level or several AI agents that each do one thing well?

This isn't a hypothetical debate. It's the exact conversation happening across staffing tech at the moment.

The generalist argument is simple: one system, one login, one thing to manage. The specialist argument is that recruitment isn't one task pretending to be simple - sourcing, screening, and placement conversations each need a different kind of judgement, and a single model asked to do all three tends to be mediocre at each rather than excellent at one.

Marika's approach is a bet on the second view. Not one AI trying to be good at everything a desk does, but several narrow agents, each built for a specific moment in the placement process, working together under one roof. It's a structural choice, not just a feature list and it's the same choice a lot of staffing businesses are quietly having to make right now as they look at what to bring in-house.

This story may be fiction but the problems are real.  Want to get the most out of your AI agents or need help building them to fit your business?  We’re here to help.


Get in touch