Screening Candidates for Recruitment: The Agency Process, End to End, and Where the Speed Comes From

Screening candidates for recruitment is the step between "this person replied" and "I would put this person in front of a client". For an agency it is also where most of the week disappears: chasing a slot for a phone screen, asking the same basics again, writing notes that nobody can compare later.

This post walks through the agency screening process end to end, shows where the time goes, and explains where the speed actually comes from. It is not about which questions to ask. If you want that, our post on knockout questions covers must-have questions and why a failed answer should be flagged, not auto-rejected.

Written in October 2026. The regulatory points below are how we read the rules, not legal advice.

What screening is for an agency

An in-house team screens applicants for its own job. An agency screens for several clients at once, and that changes two things.

First, a candidate who is wrong for today's search may be right for another client next month. A "no" from screening is rarely final. It is information about fit, for one role, at one moment.

Second, the agency is accountable twice: to the candidate, who trusted it with their details, and to the client, who expects the shortlist to hold up. A screening process has to produce answers a recruiter can stand behind in front of both.

The usual candidate screening process, and where the hours go

Most agencies run some version of the same sequence: read the CV, book a phone screen, take notes, decide, then often book a second call for whatever the first one missed.

Each step is reasonable. The cost is in the gaps and the repetition. In Totaljobs' UK research with recruiters, "It takes 3.6 hours per vacancy to filter out irrelevant applications", and "72% of recruiters say screening many irrelevant applications is slowing down the hiring process" (Totaljobs).

Look at where those hours actually go and most of them are not judgement. They are logistics: finding a time, re-asking where someone is based, typing up a call. Judgement is the part worth keeping. The rest is what a better process removes.

Speed, part one: ask the basics before anyone books a call

A phone screen is an expensive way to learn someone's location, notice period and right to work. Those are facts. They can be collected before anyone picks up a phone.

Keep that first step short. The clearest data on length comes from job applications rather than screening, so read it as an analogy. An Appcast study reported by SHRM found recruiters can lift conversion "by up to 365 percent by reducing the length of the application process to five minutes or less", and that "Completion rates drop by almost 50 percent when an application asks 50 or more questions vs. 25 or fewer questions" (SHRM).

The format matters too, especially for people who are not job hunting. LinkedIn's advice for reaching passive candidates is to "offer them an alternative way to share their background, and you probably want to do away with the application" (LinkedIn).

That is why we built PlacementFlow's screening as a web chat at a private link, not a phone screen or a long form. The people worth placing already have jobs, so a link they can open on their phone at 11pm gets their basics in without booking a call, and half-finished answers are saved.

Speed, part two: remove the waiting between steps

Candidates do not leave because a process has many steps. They leave because nothing happens between them. In Monster's survey reported by SHRM, "the top reason why candidates had withdrawn their applications was poor communication, which was cited by 47 percent of survey respondents" (SHRM). And in a 2021 Robert Half survey, "72 per cent of professionals said they lose interest in a job if they don't hear back from the employer within two weeks" of a first interview (Robert Half).

So the fastest screening process is the one with the fewest pauses where the candidate is waiting for you. For us, that meant removing the waiting: an interested reply gets the private link back in the same thread, and answers reach the recruiter as they come in.

Speed, part three: the same questions, in the same words, for everyone

This is the part most agencies skip, and it is where the recruiter's review time is won or lost.

Twenty phone screens are twenty different conversations. One candidate was asked about notice, another was not. One answer is a quote, another is a note from memory. When the recruiter sits down to decide, half the work is reconstructing what was asked.

Structure fixes that, and the evidence for it is old and solid. The US Office of Personnel Management describes the most structured interviews as "asking all applicants the same exact set of pre-defined lead and probe" questions, and notes that "Interviews with higher degrees of structure show higher levels of validity, rater reliability, rater agreement, and less adverse impact" (OPM).

PlacementFlow applies that to the first screen. Every candidate gets the questions one at a time, worded exactly as the recruiter approved them. If an answer is vague, the AI asks one follow-up, never more. The model writes the small talk around a question. It never rewrites the question itself.

The result is a set of answers a recruiter can read side by side. That comparison is where the review gets fast.

Speed, part four: one human decision point, and nothing decided by the machine

Taking the logistics out does not mean taking the recruiter out. It means the recruiter's time goes on the decision, not on the chasing.

In PlacementFlow, the recruiter reads the answers and decides. The platform flags a must-know mismatch, never rejects anyone and never sends a rejection, so every decline has a name on it.

The screen runs in two stages with the recruiter in the middle. The must-knows come first, in a short qualifier. The recruiter reviews it and decides who goes on to the deeper questions. Approving someone sends them an invitation from the recruiter's own mailbox, with the link to the next stage. Nothing moves a candidate forward or out without a person deciding it.

Using AI to screen without the traps the regulator found

If software asks the questions, two rules matter most.

Tell candidates they are talking to an AI. The EU AI Act requires AI systems that interact directly with people to be designed so that the people concerned "are informed that they are interacting with an AI system" (EU AI Act, Art. 50). That obligation applies from August 2026. Our screening chat says so up front, every time. What candidates see is described in AI Transparency: What Candidates Are Told.

Ask, collect only what you need, and never infer. When the UK Information Commissioner's Office audited AI recruitment tools in 2024, it "made almost 300 recommendations, such as ensuring personal information is processed fairly and kept to a minimum". Some tools were "inferring characteristics, including gender and ethnicity, from a candidate’s name instead of asking for this information" (ICO). A screen should ask the candidate about the things that decide the role, and nothing else.

How PlacementFlow runs it

Put together, this is the screening process we built:

A private link, not a call. The candidate opens a web chat on any device, with no login.
Must-knows first. The few facts that decide whether a conversation is worth having are asked before anything else. How to set them up is in Set Must-Knows in the Question Builder.
One question at a time, word for word. Every candidate gets the same approved wording, with at most one follow-up on a vague answer.
Every answer saved as it is given. A candidate who stops halfway has not lost their answers, and the recruiter can see how far they got.
A recruiter decides between the stages. Flags are information for that decision. Nobody is rejected automatically and nobody receives an automated rejection.

For candidates who come from a sourcing campaign, the link arrives in reply to their own email. That part is covered in How Sourcing Campaigns Work.

This is where the leverage comes from. The AI carries the operational work of screening, and the recruiter keeps every decision. That split is how one recruiter gets to roughly two to three times the operational throughput of working alone.

A checklist for your screening process

Collect the basics before anyone books a call. Location, notice and right to work are facts, not conversation.
Keep the first step short. Ask only what decides the next step.
Ask everyone the same words. Comparable answers are what make review fast.
Remove the waiting. Every pause where the candidate waits on you is where you lose them.
Keep one human decision point. Flag mismatches. Let a person decide.
Tell candidates when an AI is asking. And collect only what you need.