An agency runs two pipelines at once, clients and candidates, and the administration between them is where the margin quietly disappears.
Key takeaways
A client sends a req, usually as a paragraph in an email rather than a brief. Someone turns it into a posting and a search. Candidates come back, get formatted into the client’s preferred submission template, and get chased for availability. Interviews are coordinated between two organisations that do not share a calendar. Then there is the placement paperwork, the start-date confirmation, and the trigger for invoicing.
None of that is recruiting. All of it is the reason a recruiter carries five reqs instead of nine.
Req intake becomes structured: the client’s email is turned into a brief with the requirements, compensation range and any disclosure obligations already captured, and sent back for confirmation. Submission packs are generated in the client’s format from the candidate record, so nobody is reformatting a CV by hand at nine at night.
Interview coordination runs on the same scheduling automation an in-house team would use, except that it is bridging two calendars sets. Placement paperwork, start confirmations and the handover to the client’s onboarding process are triggered from one record rather than assembled from memory.
Ontario’s job-posting rules put the obligation on the employer, which makes it easy for an agency to assume the question is not theirs. It is. The requirement to disclose the use of AI to screen, assess or select applicants applies equally where the employer engages a third party, such as a recruiting firm, to do that screening on its behalf.
In practice that means two things. Be able to answer, in writing, exactly what your process does with AI, because clients will start asking. And when you post on a client’s behalf, treat the disclosure as part of the posting template rather than something the client remembers to ask for.
The temptation with agency automation is to point it at screening, because screening is the biggest pile. That is the one place it does not belong. An agency that lets software reject candidates has taken on its client’s human rights exposure and its own reputation risk in exchange for an afternoon.
The gain is elsewhere and it is larger: reqs per recruiter, submissions turned round the same day, and a placement process where the paperwork never becomes the reason a start date slips.
A 30-minute call is enough to tell you whether AI pays for itself here.
