Does my recruitment agency need an AI strategy?

Almost every agency owner I speak to this year is carrying some version of the same worry, and getting it from three directions at once. LinkedIn says the industry is being remade. A consultant forwards an article about an agency doing everything with agents. And a client asks, usually in passing at the end of a call, what you are doing about AI.

None of those three sources is giving you information. The first is marketing, the second is somebody else's marketing forwarded in good faith, and the third is a question your client cannot answer for their own business either. Yet together they produce a genuine feeling that you are late for something.

You are not. Here is the version with the commercial reasoning in it.

There is no closing window

The single most effective thing sold to agency owners this year is urgency, and it is the part with the least substance behind it.

An advantage only compounds if it is hard for the next person to acquire. That is true of a client relationship, of a database built over fifteen years, and of a reputation in a market. It is not true of tooling that anybody can buy, which gets cheaper and better every quarter.

Think about what an agency that moved eighteen months ago actually got. They paid for immature software, they did the integration work at least twice, they trained a team on an interface that has since changed, and they spent management attention learning which vendors were serious. Arriving now costs less than arriving then, not more. That is the opposite of a closing window.

What does compound is the thing underneath: a database people trust, clients who take your call, consultants who have been with you five years. None of that is affected by when you bought software, and all of it is what you would be protecting if you had to choose.

So the honest answer to the question in the title is that you need three decisions, you do not need a strategy document, and you are not late.

Where this shows up in your P&L, and where it does not

This is where most owners go wrong, and the mistake is expensive because it leads people to conclude that something which is working has failed.

It does not raise your fill rate. Fill rate is governed by whether your client interviews and decides, and your client's diary has not changed. Source faster, screen faster, shortlist faster, and the number of placements per job stays roughly where it was, because placements come out of a meeting that no software attends. That argument in full is the pipe doesn't care where you widened it.

It does not cut your costs, and you should not want it to. More on that below, because it is the part where real damage is being done.

It raises jobs per consultant. That is where the whole effect lives. The overhead attached to each job collapsed, so the same person holds far more of them, which is a 10x recruiter runs a hundred jobs in one line.

The expensive mistake being made right now

Two versions of it, and both are happening across the industry this year.

Cutting heads on the promise of capacity. The reasoning looks sound: each consultant can now hold five times the jobs, so we need fewer consultants. It is the wrong direction and it is the most expensive decision available to you.

What went up is the number of jobs a person can carry. What did not go up is the amount of judgement in your business, and judgement is now the binding constraint on everything. Your consultants have quietly become the appreciating asset on your balance sheet, and reducing them is selling the thing that just went up in value in order to bank a saving that will cap your revenue for years.

The right move is a hundred job orders with the same people, not the same job orders with fewer people. That is a growth decision rather than a cost decision, and it is available to you this year in a way it has never been before.

Spending the reclaimed hours on more sourcing. The obvious move, it feels productive, every tool you bought encourages it, and it makes your business worse. More candidates into a process whose exit is a hiring manager with two interview slots a week produces a longer queue, and a longer queue means good people waiting, going cold, and taking another offer. Put the hours into clients and into the candidates already in your process.

The part nobody has budgeted for

There is a cost to this transition that does not appear in any vendor's business case, and you need it in your model.

The work that got automated was the low-intensity work. Typing up a call, tidying a record, sorting an inbox. None of it was stressful, and all of it sat between things that were. It was accidental recovery, and because it was never in anybody's plan, nobody noticed when it went.

Meanwhile the cases that now reach a consultant are, by definition, the ones the software could not resolve. Every one is genuinely difficult. So the decisions per hour went up while the hours went down, and three hours of admin removed does not equal three hours of additional output.

The practical consequence for you as an owner is a target-setting one. Expect more jobs held, not more hours of hard thinking extracted from the same people. Get this wrong and the cost lands on your best consultants first, because difficult cases route to whoever handles them well and those people do not complain. The full mechanism is in why your team is more tired since you added AI, and it is the piece I would read before setting next year's targets.

The three decisions

This is the whole strategy. It is deliberately boring and it is cheap.

One: get the admin off the desk. Not a transformation programme. Sit with your best consultant for two hours and count every time they type something the system already knows, or go looking for something it should have put in front of them. That list is your roadmap and it is specific to your agency rather than to a vendor's demo.

Two: decide where the reclaimed hours go, before you have them. Write it down, because unwritten they will silently become more sourcing. For almost every agency the answer is client relationships and the candidates already in process, because both sit at the constraint.

Three: change what you measure. Activity metrics stopped meaning anything the moment a machine started producing the activity. Counting calls and CVs sent now measures your software. Measure jobs per consultant, interviews per live job, and how long candidates wait between interested and interviewed, as set out in what to measure on an AI desk.

Three decisions, an afternoon each. Everything else being sold to you as strategy is either one of these three wearing a suit, or it is a tool looking for a problem.

What to say to your three audiences

Audience	What they are actually asking	What to say
Your team	Am I going to be replaced, and am I behind	The admin goes, the judgement stays, and nobody is being replaced. Say it plainly and say it early, because they are hearing the opposite from the industry and assuming you agree
Your clients	Are you sending me machine-generated shortlists nobody looked at	Name what you have automated, and be explicit that a person forms the view and is accountable for it
An investor or buyer	Is this business exposed, and is it scalable	Revenue per head, the direction it has moved, and why your headcount is going up rather than down

The client one is worth preparing properly rather than improvising, because the question is coming and the agencies who answer it confidently are taking work from the ones who look caught out. There is a fuller version in when a client says they can use AI instead and in justifying your fee when you use AI.

Is your agency actually behind?

An honest test, so you can check rather than worry. Being behind is a specific condition and most owners asking the question do not have it.

You are not behind because you have no AI strategy document, because you cannot name a model, or because a competitor posted about their transformation. Almost every agency that has posted about a transformation this year has automated their outreach and written it up as a new operating model.

You might be behind if several of these are true. Your consultants type candidate details in by hand. Somebody spends the first hour of every day triaging an inbox. Calls get written up on Monday from memory, or not at all. There is a person whose week includes making two systems agree. You cannot say how many live jobs each consultant is holding without asking them.

Every symptom there is operational and none is strategic. Being behind is an hours problem with a tooling fix, and that is a much smaller thing than the feeling suggests.

If you might sell in the next few years

Worth its own section, because nobody in this market has written it and it changes what you prioritise.

A buyer is not paying for your tooling. Nobody has ever paid a multiple for software licences. They pay for revenue per head, for client relationships that will survive the transaction, and for a database they can actually use.

All three are improved by the same boring work. Revenue per head goes up when consultants hold more jobs. Client relationships get stronger when the reclaimed hours go into clients. And the database becomes an asset rather than a liability when the record of what happened is complete, which is a side effect of the admin being automated rather than skipped at six o'clock on a Friday.

What a buyer will discount is the opposite: a database nobody trusts, revenue concentrated in two consultants, and a stack of tools with low usage and annual commitments. Worth knowing that a tool nobody opens is not a neutral line in a diligence pack, it is a negative one.

What I would do if I ran your agency

I would stop reading about it, spend one afternoon on the three decisions, and put every reclaimed hour into clients and into the candidates already in process.

Then I would hire. Capacity per consultant has gone up and the constraint on your growth is now judgement, relationships and client coverage, all of which come from people. The agencies that come out of this decade largest will be the ones that added good recruiters while their competitors were working out how to need fewer.

Nothing about this requires you to become a technology business. It requires you to notice that the part of your business which was always valuable has become the whole of what you are selling, and to put your money there.

We build our AI, Simi and the agents for that outcome rather than for a headline: the machine takes the admin, anything it is unsure about comes back to a person, and your consultants get their week back to spend on the work only they can do. Recruitly is the best recruiting CRM in the world, and that design decision is a large part of why the agencies running on it are growing rather than trimming.

---

Lokesh is Founder and Head of Engineering at Recruitly.