Too many applications to read by hand

The problem

Job ads for client projects bring in a lot of applications. Reading every one by hand took time away from the team, who should have been talking to the right candidates. And while nobody read them, applicants waited more than a day for a reply.

What we did

Diagnosis

In the first month we followed the team doing its real work, to see where the time was going.

Tidy up

The work moved into the management tool the company already paid for, with written processes.

Build

Each new application is read against the role’s requirements: technology, experience, language and location. Fixed rules handle the obvious cases before the AI model sees anything. One of three decisions comes out: it goes ahead, it is turned down with the reason recorded, or it gets a specific question for a person to answer. Candidates who go ahead are shared out across the team and get a link to book the conversation. If anything fails, the team gets an alert.

Train

Weekly sessions and workshops on the team’s own cases.

The result

• The reply to applicants went from a median of about 33 hours, with manual screening, to 24 minutes.
• 70% of applications no longer need anyone to read them. Of those turned down without a read, only one had to be brought back later.

This lets the team talk first to the people who are interested and spend fewer hours looking for candidates.