Howdy Recruiters!
89% of HR people say AI saves them time. Only 17% say their org's AI rollout has been highly successful. Nearly a quarter have no way to measure ROI on the tools they've already bought.
So we've got high adoption, good vibes, but not much evidence of ROI.
The price of the vibe
Per-seat AI tooling adds up fast. Sourcing copilots, screening tools, scheduling assistants, summarisation layers. Stack three or four and you're looking at £150 to £300 a month per recruiter before your ATS, LinkedIn licences and job boards.
Then factor in the verification tax. The time spent checking outputs, rewriting outreach so it doesn't read like a chatbot school reunion, and chasing down the errors your hiring manager spotted before you did.
That "20% time saving" can quietly become 5% once you do the actual arithmetic. And most teams can't tell either way, because nobody clocked the baseline before the tools went live.
So what do we do Luke? Mr Doomy Gloomerson!
Three things actually....
Set the baseline before you buy
Time-to-fill, sourcing yield, recruiter throughput. Pick three you trust, write them down somewhere you'll actually find in six months. No before means no after.
Track verification time
One week. Ask your recruiters to log the minutes spent correcting AI output. That number is your real productivity figure. It tells you whether the tool is saving hours or just moving work around.
Kill what doesn't earn its seat
If a tool can't shift your baseline metrics in 90 days, cut it.
Why Everything I just said doesn't matter
The productivity gap is real. The measurement problem is real. But I'm not particularly worried about either of them.
This technology is in its early generations. It is going to get significantly better. The efficiency curve is only going one direction. The tools that feel clunky and expensive to verify today are going to look primitive in three years.
The recruiters and TA leaders who come out of this period well aren't going to be the ones who waited for proof before they engaged. They're going to be the ones who got stuck in now. Who ran the experiments, shared what they found, built the muscle, deployed imperfectly, and kept iterating.
There's an upfront efficiency cost to learning anything new. AI is no different. The people building a moat around "we don't need this yet" are storing up a much bigger problem for later.
So yes, set your baselines. Track your verification time. Cut the tools that don't pull their weight. Do the maths properly.
But do it as someone who is actively in the game, not standing at the side waiting for the results to get tidier.
That's where the compounding happens.
Big tip of the cap to Hung Lee and the Brainfood team for the source material. If you don't subscribe, you're missing the best curation in the industry.
Luke Update
Week of May 5th I'll be in San Fransisco for the AshbyOne conference. If you're kicking about, hit me up! We can go look at bridges and eat seafood or... something
Well... that's issue seventy! If you have any questions about it, or any feedback on this issue of The Data Driven Recruiter, grab me on LinkedIn for a chat.
I'll see you next week!