The Candidates You Already Have

A role comes in. Somebody posts it, buys a job board slot, and starts sourcing.

Meanwhile the database holds four thousand people who applied for something over the last five years, and a good number of them are right for this role. Almost none of them will be contacted, because finding them would take longer than starting fresh.

The asset you paid for and never used

Every one of those records cost money. Advertising, screening time, an interview, a conversation. That spend is sunk and the value was never collected, because the person was not right for the role that week.

Being wrong for one role in 2022 says almost nothing about being right for a different one now. People move, get promoted, finish qualifications, relocate, and change what they want. A database is not a list of rejections. It is a list of relationships that were never maintained.

Why nobody works it

Because searching it properly is genuinely hard, and everybody who has tried knows it.

Keyword search returns the people who happened to use your search term in their CV. It misses the candidate who wrote “site supervisor” when you searched “foreman”, the one whose relevant experience is described in a paragraph rather than a job title, and the one whose CV is a PDF that never got parsed properly.

So consultants search, get twelve mediocre results, conclude the database is useless, and go to the job boards. The database is not useless. It is unsearchable by the methods available, which is a different problem.

The second reason is that the data decays. A CV from 2022 describes somebody who no longer exists professionally, and nobody has time to ring four thousand people to find out what changed.

What AI does on a recruitment desk

This is one of the strongest cases in any industry, because the entire problem is reading unstructured text and making a judgement about fit, at a volume no person can reach.

It reads the CV rather than scanning it for keywords. A person who spent four years running a crew of twelve is a foreman whatever their job title said, and that is now an obvious match rather than an invisible one.

It searches on the brief, not on terms. Give it the actual role, in the words the client used, and it returns the thirty people worth a conversation with a reason attached to each. What used to be a morning is now a minute, and it is better.

It infers what changed. Somebody who was a junior in 2021 is not a junior now. A candidate finishing a two year qualification then is qualified today. Those are obvious inferences that no keyword search will ever make.

It keeps the database alive. Periodic contact with every record, written from what that person actually does rather than a mass email, asking what has changed and updating what comes back. That is the job nobody has ever had time to do, and it is the difference between an archive and an asset.

Speed is the whole business

In staffing, the first agency to present a strong shortlist usually wins the placement. Everything about the economics follows from that.

Working the database is the only route to a shortlist in two hours rather than two days, because sourcing new candidates has a floor: people have to see the advert, decide, apply, and be screened. That is days, and days are how placements are lost.

The agencies that consistently place first are not working harder. They are starting from a list they already own.

The contact has to be worth receiving

A warning, because this is where reactivation campaigns go wrong and burn the asset.

A mass email to four thousand people saying you have exciting opportunities produces unsubscribes and does real damage. These are people who spoke to you once and heard nothing for three years. The first message back has to be specific: this role, this reason you thought of them, this is what you remembered about their situation.

That level of personalisation across four thousand records is exactly what was impossible before and is straightforward now. It is also the entire difference between reactivation working and it costing you a database.

The client side of the same asset

Everything above applies equally to the other half of a recruitment business, and it is the half that usually gets less attention.

Your database holds every company that ever briefed you, including the ones who used you twice in 2021 and went quiet. Same problem, same cause: contacting them usefully requires knowing what they hired for, who left, and what has changed at that business since. Nobody has time to reconstruct that for two hundred dormant clients.

An AI system reading your placement history and the public record can. A company that has posted four roles you were not asked about is a specific, timely conversation rather than a cold call, and it is the sort of thing that sits in plain sight and gets missed because nobody is looking.

What to keep human

Two things, and they are the two that decide whether the placement holds.

The conversation with the candidate about whether they actually want the role, which is where an experienced consultant hears the hesitation that no CV contains. And the read on the client’s culture, which determines whether a technically perfect candidate lasts six months.

AI gets you to a strong shortlist in an hour instead of two days. It does not make the judgement about people, and a desk that tries to hand it that job will place faster and lose more of them at eight weeks.

Do this before you buy anything

• Count the records and count how many were contacted in the last twelve months. The ratio is usually under five percent.

• Take your last five placements and check whether a suitable candidate was already in the database when the role came in. In most agencies it is three of the five.

• Work out what those placements cost to source against what they would have cost from the database.

That third number is the argument. It is normally large, and it is money you have already spent once.

AI Optimize makes your existing database searchable on the brief rather than on keywords, and keeps it current without anybody making four thousand calls. That work sits under Custom AI Integrations.