Your recruitment agency tech stack: what to buy, and what to cut

Most recruitment agency tech stacks are not designed, they are accumulated. An ATS in year one, a sourcing tool when a recruiter complained about candidate volume, a sales platform with the new BD hire, two AI subscriptions because a competitor mentioned them on LinkedIn. Nobody ever removed anything. Now you pay four figures a month for software your team half uses, your data sits in three places, and none of it is making anyone bill more. I have built these stacks twice from zero and inherited other people's more than twice. The problem is almost never that the firm bought bad software. It is that the firm bought software before deciding what the software was supposed to do.

Here is how I think about a recruitment tech stack: what it has to do, what it should cost, where AI actually earns its keep, and how to find the money you are already wasting.

What does a recruitment agency tech stack actually need to do?

Four jobs. That is the whole list, and if a tool does not clearly own one of them, it is a candidate for cancellation.

One, a system of record. Your ATS and CRM: where every candidate, contact, company, role, and interaction lives. This is the only tool you truly cannot compromise on, because it is the asset. When you eventually sell the firm, or lose a biller, the value that stays behind is the data in this system and nowhere else.

Two, finding people and their contact details. Sourcing and contact data: a search layer over profiles plus verified emails and mobile numbers. Most firms overbuy here and treat every provider as additive, when in practice two good sources cover the overwhelming majority of what your niche needs.

Three, getting a message in front of them. Outreach and sequencing across email, phone, and messaging, with the activity logged back into the system of record automatically. If a recruiter has to copy anything by hand, it will eventually not get copied.

Four, getting paid. Back office: invoicing and accounting for a perm book, and timesheets, pay, bill, and compliance if you run contract. Contract firms who bolt this on late pay for it in payroll errors and cash flow they cannot see. I covered why that cash discipline matters in contract vs perm staffing.

Everything else, and I mean everything, is optional until you can name the job it does and the tool it replaces.

How much should a recruitment firm spend on software?

Set a ceiling before you shop, or the stack expands to fill whatever your bank balance allows. Total software spend should be a small, deliberate percentage of gross profit, with every seat justified against the fee it helps produce. If a recruiter carries a few thousand a year in tooling, that has to be trivial against what the desk bills. When it is not trivial, you do not have a software problem, you have a productivity problem, and buying more software makes it worse.

Two rules. Price the stack per producing recruiter, not per employee: support staff rarely need the expensive seats, and vendors will happily sell you firm-wide licences. And never sign a multi-year deal in your first two years in a category. You do not yet know what you need, and the lock-in discount is worth far less than the freedom to walk. If you are still building the launch budget, that number lives inside recruitment agency startup costs.

How do you choose the right ATS or CRM?

Choose for adoption, not for features. Every platform in this market has a feature list long enough to justify the price on a slide. What actually separates them in daily use is whether your recruiters will log activity in it without being chased. A mediocre system your team uses religiously beats a best-in-class system half your desk works around, because a system of record with gaps in it is not a record.

Three things I press hard on before a firm commits:

• Is it genuinely a CRM as well as an ATS? Recruitment is two businesses running at once, candidate supply and client demand, and tools built only for candidate tracking leave your BD pipeline stranded in a spreadsheet. That is how firms lose track of the client relationships that generate the fees.
• Can you get your data out? Ask, in writing, before you sign: what does a full export look like, in what format, and what does it cost? Vendors who get vague here are telling you exactly how the relationship ends.
• What does the integration layer look like? Your ATS is the hub. If your outreach tool, your job boards, your calendar, and your back office cannot write into it cleanly, your recruiters become the integration layer, and they are expensive middleware.

Then commit. The most expensive habit a growing firm has is migrating systems every couple of years because each new tool looked better in a demo. Every migration costs you dirty data, weeks of adoption, and a quiet dip in billings nobody attributes to the switch.

Where does AI actually earn its cost in a recruiting stack?

I build AI-enabled workflows for the firms I work with, so take this as a practitioner's split rather than skepticism. AI is worth real money where the task is language work with a human check at the end.

What genuinely pays for itself today: call and interview summarisation that writes structured notes straight into the ATS, which is the highest-value use in most firms because recruiters hate writing notes and your data quality dies without them. Turning a messy job description into search strings, screening questions, and a candidate brief in one pass. First drafts of outreach and of client submission summaries. Cleaning, deduplicating, and enriching records at volume. Every one of these attacks admin time, which is the largest silent cost on a desk.

What does not yet deserve blind trust: autonomous sourcing and "match scores" presented as judgement. The models are good at retrieving people who look like a description and unreliable at the thing your fee depends on, which is knowing who will actually move, for what, and why. Treat AI output as a first pass to be worked, never as a shortlist to be sent.

The mistake to avoid is buying AI as a separate product category. The value is your existing workflow doing less manual work, not another tab nobody opens. If a recruiter has to leave the system of record to use it, adoption is near zero inside a month.

What should you cut from your stack right now?

Run this audit this quarter. It takes an afternoon and it usually funds something better.

Pull the login data. Every serious vendor can show seat-level usage. Any seat with no meaningful activity in 30 days gets cancelled or reassigned, no debate, no "but I use it occasionally." Occasional use is what one shared licence is for.

Map tools against the four jobs. Write the four jobs across a page and put every subscription under one of them. Anything you cannot place is shelfware. Anything with three tools stacked under it is duplication you pay for twice.

Find your auto-renewals. Pull twelve months of bank and card statements and list every recurring software charge with its renewal date and notice period. Almost every firm I look at finds a contract that quietly renewed for a tool nobody has opened since its champion left.

Then renegotiate rather than churn. Take real usage data into renewal calls. Vendors discount to retain accounts far more readily than owners assume, especially when you can show exactly how many seats you are about to drop.

How do you know the stack is actually working?

Measure adoption and data quality, not feature usage. Three numbers tell you almost everything.

Capture rate: what proportion of real client and candidate conversations end up logged in the system of record. If your recruiters are having conversations the CRM never sees, you are running a firm on personal relationships you do not own. That is the risk that shows up the day a biller resigns.

Admin hours per recruiter per week: estimate it honestly, then track it after every tooling change. The entire point of the stack is to convert admin time into candidate and client time. If the hours have not moved, the tool has not worked, regardless of how good the demo was.

Time from role open to first candidate contact: the cleanest measure of whether your tools actually accelerate the workflow rather than just documenting it. Speed is what clients experience, and it is the metric behind most of what I wrote in how to increase recruiter productivity.

Buy the workflow, not the software

The best stack I have run was not the most expensive. One system of record everybody actually used, two sourcing sources instead of five, outreach that wrote itself back into the CRM, clean back office, and AI pointed squarely at the admin eating my recruiters' selling time. Everything else got cut, and billings went up, because the team spent the day on the phone instead of in tabs. Decide what the work should look like, then buy the smallest set of tools that makes that work happen. If the tech is deciding your process, you bought backwards. For where those savings land, read recruitment desk economics.