A Listed Recruiter Just Blamed AI for Falling Fees - In Writing

On 16 June 2026, the STEM staffing specialist SThree told its investors that net fees for the half year had fallen to 147.7 million pounds - down 7% on the same period a year earlier. For a recruitment business, net fees are the number that matters. It is the money left after pass-through costs, the closest thing the industry has to true revenue. And it is shrinking.

That alone is not news. Recruiters have been reporting soft numbers for two years. What made this result different was the reason the company gave for it.

The line that changes the story

SThree said its clients are "reassessing staffing needs as they seek cost savings and productivity gains from artificial intelligence." Read that slowly. This is not a journalist's interpretation or an analyst's theory. It is a publicly listed recruitment firm explaining its own falling numbers to the people who own it, in a regulated market update, and naming AI as part of the cause.

For two years the prevailing explanation for the recruitment downturn was cyclical. Interest rates were high. Confidence was low. Hiring had paused. Cyclical problems share one comforting feature: they end. The market turns, confidence returns, and the paused roles come back. Every recruiter who lived through 2009 or 2020 knows the pattern - you survive the trough and the volume returns.

What SThree described is not cyclical. It is structural. Clients are not pausing roles until conditions improve. They are removing roles because AI now does part of the work those roles existed to do. And a role that has been repriced or removed for a structural reason does not return when confidence does.

A paused role comes back. A repriced one does not.

This is the distinction that should keep agency owners awake. A paused requisition is deferred demand - it sits in a client's plan, waiting for a green light. A repriced function is destroyed demand. When a client decides that a team of ten can now do the work of fifteen because software handles the rest, the five roles are not paused. They are gone from the plan entirely, and so is the placement fee that came with filling them.

The numbers underneath SThree's result show the unevenness of it. Germany net fees fell 14%. The Netherlands fell 24%. The United States, meanwhile, grew 12%. This is not a uniform tide going out. It is specific markets and specific kinds of work being hollowed out faster than others, while other parts of the business still grow. The danger for an owner is averaging it all together and concluding "the market is just soft" when in fact one part of the book is being structurally repriced while another holds.

Why the framing matters more than the percentage

A 7% decline is recoverable. SThree itself framed the result with "cautious optimism" and noted the rate of decline had improved from the previous quarter. The point of paying attention here is not the size of the drop. It is the appearance, for the first time, of AI as a named cause in a financial filing.

Listed companies do not put causes in official results lightly. The language is reviewed by lawyers and investor-relations teams precisely because the market acts on it. When a board signs off on "clients seeking productivity gains from artificial intelligence" as an explanation for falling fees, it is making a considered statement that this is a real and continuing factor - not a one-off, not an excuse, but a trend management expects to persist.

The agencies most exposed to this are not the ones with the worst quarter. They are the ones telling themselves the same story SThree told for the previous two years - that the softness is the economy, that it will pass, that no fundamental change to the offer is required. SThree blamed the economy too, right up until it had to write down what was actually happening.

What the data says about which models hold

The geographic split inside SThree's results is a clue. The US grew while parts of Europe collapsed. Across the listed recruiters, a pattern has been visible all year: specialist, contract-heavy, deeply-nicheed books have held up better than generalist, perm-heavy, volume books. The reason is straightforward. The work AI compresses first is the standardised, repeatable, high-volume kind - the exact work a generalist perm desk runs on. The work that resists automation longest is the specialised, relationship-heavy, judgement-dependent kind.

This is the same lesson the contrast between high-performing and collapsing recruiters has been teaching all year. When two firms operate in the same market under the same conditions and one grows while the other shrinks, the difference is rarely effort. It is the structure of the book - how much of the fee is attached to work a machine can now do, and how much is attached to judgement it cannot.

The slow drift is the real danger

Nobody decides to halve their business. It happens a few percentage points at a time, each quarter explained away as the market, until one day the model is half the size it was and no single decision made it so. That is the genuine threat in the SThree result - not a dramatic collapse, but a quiet, compounding drift that is easy to rationalise and fatal to ignore.

The agencies that hold their fees through this are not the ones waiting for hiring to bounce. They are the ones who have already moved the centre of their value to the part a client cannot automate away: the judgement on who to hire, the read on a candidate that no benchmark captures, the honest conversation that tells a client the role they have written is the wrong role. The labour of finding people is being automated. The judgement around the hire is not.

The honest question for owners

The uncomfortable exercise is to look at your own fee income and ask the SThree question of it. Is it soft because the market is down and will recover - or because the work behind it is being quietly automated, and the recovery will arrive without the fees attached? The answer determines whether you wait, or whether you move. SThree's mistake was waiting until the cause had to be written down. The agencies that act before that point are the ones that still have a fee to defend.