A flight-risk pattern shows up in the CV before it shows up in a reference check: three roles in four years, each lasting under 14 months, with no obvious upward trajectory. AI screening can flag that pattern across 200 CVs in minutes — the judgment call on whether it actually matters still belongs to you.
It's Monday morning. You've got 200 CVs against a mid-level ops role, a client expecting a shortlist by Wednesday, and three other live briefs pulling at the same desk hours. The client mentioned, almost in passing, that the last person they hired through another agency left after eight months — so this time they want "someone who'll stick." That's not a line item on the job spec. It's a filter you're expected to apply anyway, on top of the skills match, while reading CVs at the pace the week demands.
Where the current process breaks down
Tenure patterns are exactly the kind of signal manual screening is bad at catching consistently. At 90 seconds a CV — realistic for a first pass across 200 applications — you're skimming job titles and dates, not doing the maths on average tenure per candidate. By CV 120, fatigue sets in and the read gets shallower, not more careful. A candidate with four jobs in five years might get flagged if you're fresh at 9am and waved through if you're on autopilot at 3pm.
The other failure mode is context-blindness. A short stint reads very differently depending on why it happened — a redundancy round, a fixed-term contract that ran its course, a company that folded, versus someone who leaves every role at the first sign of friction. Scanning dates on a CV doesn't tell you which. Recruiters end up either over-flagging (penalising candidates for circumstances outside their control) or under-flagging (missing a genuine pattern because one CV among 200 doesn't get the scrutiny it needs). Neither serves the client, and neither protects your placement fee if the hire doesn't last past probation.
A workflow that catches the pattern without slowing you down
The fix isn't to screen more carefully — it's to let a tool do the consistent, structural pass so you can spend your attention on the calls that actually need judgement. Here's a workflow that works for a high-volume brief like this:
1. Set the tenure baseline before you screen. Ask the client what "stable" means for this role — some functions (contracting, project-based delivery) have naturally shorter average tenures than others. A blanket "no job-hopping" rule misreads entire sectors.
2. Run the longlist through AI screening for a first pass. This is where a tool like CV Matcher fits into the workflow — it can process the full CV pile against the brief and surface tenure patterns (average time per role, frequency of moves, gaps) alongside the usual skills and experience match, rather than you calculating it by hand across 200 documents.
3. Sort flagged candidates into a separate lane, not a reject pile. A short-stint pattern is a question to ask, not a disqualifier. Treating it as an automatic cut loses candidates who left a bad situation for good reasons.
4. Build your shortlist from skills match first, tenure pattern second. Don't let the flight-risk flag override a strong skills fit — it's one input, weighted against everything else the brief asks for.
5. Carry the flag into the qualification call as a specific question, not a vague concern. "I noticed three roles in the last four years averaging around a year each — walk me through what happened at each of those" gets you a real answer.
What the algorithm can't tell you
This is where the qualification call earns its place in the process. AI screening can tell you a candidate changed jobs four times in five years. It can't tell you that two of those moves were the same employer restructuring twice, that the candidate turned down a counter-offer to leave a toxic manager, or that they're now explicitly looking for a role they can commit to for the next chapter of their career because they've just bought a house nearby. A 20-minute call surfaces motivation, and motivation is what actually predicts whether someone stays — not the raw count of roles on a CV.
It also lets you correct for pattern-matching that's too blunt. Some of the most stable hires you'll ever place had a rocky-looking CV for reasons that had nothing to do with their commitment. The flag is a prompt to ask a better question, not a verdict.
What to try on your next brief
On the next high-volume role, before you open the CV pile, agree the tenure baseline with your client in the same conversation where you agree the skills brief. Then run the longlist through AI screening rather than eyeballing dates yourself — you'll get a consistent read across all 200 CVs instead of a read that degrades as the day wears on. Use the flagged patterns as call prep, not a cut line, and you'll walk into Wednesday's shortlist meeting with answers to the stability question before the client has to ask it. If you want to see how this fits your own screening process, you can Start Free Trial and run it against your next live brief.