Agencies and outstaff: a shortlist you can stand behind

You send three candidates. The client interviews all three and passes on all three. The feedback, when it arrives, is one line: not quite what we are looking for.

Now you are in the worst position in this business. You have spent the sourcing, you have spent the client's interview slots, and you have learned almost nothing about what to send next. Meanwhile the client has learned something about you, and it is not the thing you wanted them to learn.

This page is about the step that sits between sourcing and submission, and about what it is worth to make that step something you can point at.

What an agency actually sells

Not access to candidates. The client can post the job themselves, and in 2026 they will get more applications than they can read.

What an agency sells is judgement: the claim that these three, out of the hundred who were available, are the ones worth an hour of the client's time. Everything else is logistics.

Which makes the screening step the product. And the screening step, in most agencies, is a phone call and a recruiter's read of a CV, both of which are real skills and neither of which produces anything the client can inspect. The client sees the output, three names, and has to take the reasoning on trust.

That works when the relationship is old and the placement record is good. It works badly when you are new to the account, when the role is technical and your recruiter is not, and when the last two submissions did not convert.

Bullhorn's 2026 GRID Industry Trends Report found that firms which submit multiple candidates to a position, and those with a redeployment plan, are much more likely to report time to place under twenty days, with the redeployment figure at 85% of firms. Read alongside the rest of that report, the pattern is not "send more" but "know your bench and your pipeline well enough to move fast without guessing". Screening is what turns a list of names into a bench you know something about.

The artifact you screen on stopped being reliable

The old workflow assumed the CV carried information. Skim it, spot the mismatch, make a judgement call, phone the ones who look plausible.

That assumption is now weak in both directions. ZipRecruiter's 2026 AI Employer Report found that just under half of employers say AI has increased application volume per role, and that only about a quarter believe they can almost always tell when AI was used in an application. A tailored, competent CV is now available to anyone who wants one, including candidates who cannot do the job and candidates who can but would never have written it that way themselves.

For an agency this is worse than for an employer, because the CV was your raw material. When the raw material stops separating people, the phone call has to do all the work, and the phone call is your most expensive resource per candidate.

The move: a written screen before submission

One questionnaire per role, or per client, written once. Every candidate you are considering answers the same questions in writing, in their own time, before you decide who goes forward.

What that gives you is three things the phone call does not.

A comparison instead of a sequence. Phone screens happen one at a time, days apart, and the fourth candidate is compared against a memory of the first. Written answers sit next to each other.

Evidence you can quote. When you tell the client why this person, you can point at what the candidate actually wrote about the exact problem the client cares about. That is a different conversation from "she interviewed well".

Coverage of the technical ground your recruiter cannot cover. If the client needs someone who has run a migration under load, you do not need a recruiter who can assess that. You need the question asked, the answer recorded, and a report that tells you which competencies the answer actually demonstrated.

Where the questions come from

For agency work the questions should come from the client, or at least be approved by them. That is not a formality, it is the whole mechanism.

A short call at the start of a role, aimed at four questions rather than at a job description, is one of the highest-leverage conversations in this business. It forces the client to say what they will actually reject people for, which is the thing job descriptions systematically fail to contain.

Then the questionnaire becomes reusable in two directions. Across candidates for that role, obviously. And across roles for that client, because the second role from the same team usually needs three of the same four competencies. In Eneya a questionnaire can be duplicated and edited rather than rewritten, which is the difference between doing this once per client and doing it once per search.

There is a side effect worth naming. When you send the client the questions you intend to ask before you ask them, you are showing your working before there is anything at stake. New accounts in particular respond to that, because it turns "trust our judgement" into "here is our judgement, in advance, tell us if it is wrong".

A note on who answers first. In agency work the candidate is often not yet committed, and the order of steps matters more than in in-house hiring. Sending a questionnaire before you have spoken to someone at all reads as a filter applied by a stranger. Sending it after a short conversation, framed as the thing that replaces the client's generic first call, reads as representation. Same questionnaire, different placement in the sequence, very different response rate.

What you send the client

Be careful here, because this is where it is easiest to promise something the tooling does not do.

The scored report lives in your account. Your client does not log in to read it, and there is no public link to hand them. What you send is what you always sent: your own submission note, a CV, and your recommendation. What changes is what stands behind it.

Concretely, that means you can:

• Quote the candidate's own words on the specific competency the client cares about, rather than paraphrasing an impression from a call.
• Say what was asked, which is often more persuasive than the answer. A client who sees that you asked about failure modes under load knows what kind of screen this was.
• Export the list of candidates and outcomes as a CSV when you want a summary of a search, with the verdict and your decision as separate columns.

What you should not do is imply to the client that they will get a portal, a dashboard, or a shareable report link, because they will not. If the client wants the underlying evidence, you send them the text of the answers, which you have.

Outstaff is the same problem with a harder bar

Outstaffing and body-leasing carry a version of this that bites harder. The person you place will sit inside the client's team, be reviewed by the client's engineers, and reflect on you every day for the length of the contract. The client's technical bar is the bar, and you do not have their engineers to help you screen.

Two things follow.

The questionnaire is the transfer mechanism for the client's bar. Their competencies, their scenarios, their idea of what senior means. Written down once, applied identically to everyone you propose. It is much easier to argue with a client about four written questions than about a rejection after the fact.

A screen that produced a placement is worth more than one that produced a rejection. After a placement lands, go back and look at what that person wrote. That is the closest thing to a calibration set an agency can get without building an assessment function, and it costs nothing except the habit of looking.

One honest caveat about reuse. Candidate answers are personal data, and Eneya deletes them on a retention schedule tied to the hiring process rather than keeping them indefinitely, with the exact periods in the privacy policy. So this is a way to know your bench during a search and shortly after it, not a permanent talent database. If you want a long-lived record, what belongs in your ATS is your own note about the person, not a copy of their answers.

The two objections you will hear

From candidates: another step? Yes, and it has to be small enough to be worth it. Three or four questions, a stated time, and a clear reason: this is what the client wants to know, and answering it means you skip the generic first call. Candidates who are being represented, rather than processed, can tell the difference. Bullhorn's 2025 GRID Talent Trends Report found that candidates rank the right opportunities and fast placement above almost everything else, and that satisfaction with speed and responsiveness has been falling. A short screen that visibly shortens the rest of the process is defensible on exactly those grounds. A long one is not.

From the client: just send us CVs. Some will say this, especially when they are in a hurry. The answer is not to argue about process. Run the screen anyway, send the same three names you would have sent, and put one quoted answer in the submission note. The next role from that client tends to arrive with the questions attached.

The AI question, from the agency side

Employers are increasingly wary that the material in front of them was generated, and agencies inherit that suspicion twice: once for the candidate's CV, and once for the submission note if it reads like everyone else's.

The candidate half is manageable and mostly a matter of question design. Questions that ask what the person did, on what system, and what went wrong are much harder to answer from a prompt than questions that ask what they think about a topic. On top of that, the answers either name something a client could check, a company, a product, a project with details tied to it, or they do not. Eneya reports that as a High or Medium reading next to the score, with the fragments that gave it, and turns any two answers that disagree into questions for the call. It does not guess whether a model was involved, and neither should the submission note.

Treat it as a reason to ask, never as a verdict. A fast, tidy answer often means the candidate has explained this exact thing in five interviews already, which at agency volume is the common case rather than the exception. The follow-up costs one message: ask them to walk through the specific instance they described.

The second half matters more for your business than the first. A submission note that could have been written about anyone is now the norm, because it is free to produce. A note containing two sentences the candidate actually wrote about the client's exact problem cannot be produced without doing the work, and clients notice the difference long before they can articulate it.

What the client is really asking when they ask why

"Not quite what we are looking for" is not obstruction. Usually it means the client cannot articulate the gap either, and is hoping you will figure it out.

A written screen turns that into a tractable conversation, because you have something concrete to put in front of them. You can ask whether the answer to question two was the problem, and whether the bar on that competency is higher than they described. Half the time the client discovers their own requirement while reading a candidate's answer to it.

That is worth more than the placement it saves. A client who has been walked through their own bar once will describe the next role better, and the next search starts closer to the target. This is the compounding part of the whole approach, and it does not come from the scoring. It comes from having asked everyone the same questions and kept the answers.

What this does not replace

It does not replace the client's own interview, and no agency should want it to. The client owning the final decision is what keeps the placement theirs.

It does not replace references, trials, or probation. For contract and outstaff work especially, the first two weeks remain the real assessment.

It does not replace your relationship with the candidate. A questionnaire is not a substitute for knowing what someone wants next, what they will not take, and what would make them leave. That is still the part of the job that is actually yours.

And it does not make the decision. The score is computed from the model's assessments by fixed code in the product, and it is advisory. Your recommendation to the client is a judgement you make, and it should stay that way, because it is the thing you are paid for.

Where this breaks down

Retained search with three candidates. If the whole longlist is five people and you know all of them, this adds a step and subtracts nothing. Just talk to them.

Very senior roles. For a director or a CTO, the screening question is fit, politics and appetite, and none of those survive contact with a questionnaire. Written screening lives in the layer below.

Speed-only markets. In some volume contract markets the first submission within an hour wins, and a step that takes candidate time is a losing trade. Know which of your accounts work that way, and do not apply the same process to both kinds.

Clients who will not name a competency. If the client cannot tell you what they will reject for, the questionnaire will be generic, and a generic questionnaire produces a confident number about nothing. Sometimes the honest outcome of that call is that you are not ready to screen yet.

Starting this week

1. Pick one account where the last submission did not convert, and where you would like to know why.
2. Ask the client for four things they would reject a candidate for. Write one question per item. Half an hour on a call, and it is a better call than the one you were going to have.
3. Send the questionnaire to your current longlist for that role, including the two people you were unsure about.
4. Form your own view of each answer before you look at any score. Then submit as you normally would, with one quoted answer in the note.
5. When the client responds, compare their reaction to your view and to the score. Three roles of that and you will know how much either one is worth on that account.

The pitch to the client never changes. You are selling judgement. This is a way to have some of it written down before anyone asks you to defend it.

If you also hire engineers directly, the same idea applied to protecting senior interview time is on the engineering hiring page. If your problem is the size of the pile rather than the client relationship, start with high-volume roles.