Use AI screening when the role has criteria you can write down and a pile of CVs too big to read properly. Go manual when the real brief lives in the client's head, or when the search can't be advertised. Two questions decide it: how many CVs are coming in, and how much of "good" can you put on paper?
Picture Monday morning. You have 200 applications for a finance operations role, the client wants a shortlist by Wednesday, and two other briefs are waiting for desk time. On the same day, a board member rings about a confidential CFO replacement that nobody inside the business knows about yet. Both are "screening", but they need different tools.
Where the one-size-fits-all approach breaks down
Treat every role the same and one of them suffers. The maths on volume roles is simple: 90 seconds per CV across 200 CVs is five hours of reading before you've made a single qualification call. That's a full working day lost to first-pass sorting, and it assumes your attention holds up. It doesn't. CV number 140, read at 4:45pm on a Monday, doesn't get the same scrutiny as CV number 12. Fatigue makes your screening inconsistent, and the candidate who gets rejected unfairly is usually the one with an unusual layout or a non-linear career path.
The reverse problem hits executive searches. Run a senior, confidential brief through a volume workflow and you create risks that no time saving justifies. Sensitive details end up in more systems than they need to. A "match score" flattens someone whose value is a track record of turning around a failing division, which no job title captures. And the shortlist is only fifteen people long anyway, so there's no volume problem to solve.
The failure mode is the same in both directions: the method doesn't fit the shape of the brief. Volume roles need speed and consistency. Executive roles need discretion and depth.
A simple way to sort your roles
Score each incoming brief on four questions before deciding how to screen it.
1. How many CVs will you realistically see? Past about 50, consistency starts to slip. Past 150, you're triaging rather than screening. Under 20, reading them yourself costs very little.
2. Can you list the must-haves in a few lines? "Three years in AP/AR, month-end close experience, SAP" is machine-checkable. "Someone who can win over a sceptical founder-led board" is not.
3. Is confidentiality a condition of the search? If the client's name, the incumbent's situation or the reason for the vacancy can't leave a small circle, keep it off any platform and out of any shared document.
4. How much does culture and context decide the outcome? The more the placement depends on reading the room, the less a first-pass tool can tell you.
Roles that score high on volume and clarity, and low on confidentiality and culture, are your AI candidates. Customer operations, sales development, finance and accounting staff, IT support, logistics and most mid-level technical roles usually fit. For these, a tool like CV Matcher reads each CV against the brief and ranks candidates by fit. It's semantic rather than keyword-based, so it can catch the person who did the work under a different job title, which is the kind of match keyword ATS filters miss. You then spend your hours on the top of the ranked list, not on the first 40 CVs that happened to arrive.
The quality of that ranking depends on the quality of your input, so spend ten minutes on the brief first. We cover how in structuring your job brief so AI screening works. Separate hard requirements from preferences, and say what a "no" looks like.
Roles that score high on confidentiality and culture, and low on volume, stay manual: C-suite, board-adjacent, founder-successor and any search where a leak would damage the client or the candidate. You're working from a target list, referrals and direct approaches, and you're the one reading the career history.
Many roles land in the middle, such as a head of department with 60 applicants. A hybrid works well there: let the tool do an initial sort of the non-sensitive material, then read the top group yourself, in full.
What only you can tell
Even on the most automatable role, a ranking is a starting point. It can tell you who matches the brief on paper. It can't tell you why someone left three roles in four years, or whether that was a pattern or three bad employers. (Our piece on spotting flight-risk candidates before you shortlist goes deeper on that one.)
A 20-minute qualification call reveals things no algorithm measures: whether a candidate's reason for moving is credible, how they talk about a former boss, whether their salary expectation matches what they'll actually accept, and whether they'd be comfortable with this client's way of working. In executive search, that conversation is most of the job. You're judging presence, how someone handles a hard question about a failure, and whether they'll still be in the seat after the first difficult board meeting.
AI narrows the field. You decide who gets put in front of your client.
Your first step on the next brief
When the next role comes in, run the four-question check before you open a single CV. Write down the volume estimate, the must-haves, any confidentiality constraints and the culture weighting. It takes five minutes, and it settles your method before the pile builds up.
If the role is a high-volume, clear-criteria one, test the AI route on it. You can Start Free Trial and run a live brief through it. Compare the ranked top 15 against the 15 you'd have picked yourself. Where you disagree, work out whether the tool or your gut was right. Then keep the executive searches firmly in your own hands.