The agency runs four desks against a shared applications inbox. Every morning each consultant opened the same pile, worked out which applications were new, which were the same person applying a third time under a different address, and which had nothing useful attached. That triage was the first hour of the day, every day, and it was the part of the job nobody had been hired to do.
What was actually going wrong
The problem was not volume. It was that the useful work sat behind an hour of sorting that could not be skipped, and that the sorting was done four times over because nobody could see what the other desks had already handled.
- The same candidate arrived under a personal address and a work address and counted as two people.
- Applications with no CV attached looked identical to complete ones until opened.
- Nobody could tell whether a name had already been screened by another desk that week.
- Enrichment, when it happened, was one consultant pasting a name into a search box.
What we put on the board
Three seats, working overnight against the inbox and the applicant tracking system. The Record Cleaner matched incoming applications against existing records and merged the duplicates. The Account Researcher pulled a short public profile for each new name and attached it to the record. The Support Triager sorted what was left into complete, incomplete and out of scope, and left the incomplete ones a note about what was missing.
Everything ran on watch only for the first two weeks. The consultants could see each decision in the run trace and undo it, and the seats did not touch the tracking system until the team had read a fortnight of runs and agreed with them.
How the day works now
The consultants open a sorted queue. Duplicates are already merged, each candidate carries a short profile, and anything missing a CV has already been chased. The hour that used to go on triage goes on calls.
Nobody on the desk was hired to deduplicate an inbox, and now nobody does.
What it cost
Every paid action is checked against a cap before it runs, so the monthly spend is a number the agency set rather than a number they discovered. The run trace shows the cost of each run beside what it did.