Screening 10,000 CVs a week with AI

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

The challenge had three focus areas.

Each incoming CV is passed to a workflow built on Amazon Bedrock. A large language model reads the CV alongside the structured job specification for the role, which sets out the required experience, the ideal candidate character, the package and the client's success criteria. The model evaluates how well the candidate matches and returns a classification with a short rationale.

The proof of concept demonstrated that a Bedrock-based workflow could read a real CV against a real job specification and return a sensible, structured classification, using anonymised example CVs and a live role specification as test data. It showed a credible path to automating the weekly triage burden while keeping a human in the loop for borderline cases.

• Automated triage proven — A Bedrock workflow classified CVs against a structured job specification, tested with anonymised example CVs and a live role.
• Human in the loop — The middle "debate" group keeps a consultant in control of the borderline decisions where judgement matters most.
• Works with existing tools — Classifications flow into Salesforce, so the tool speeds up the team's current pipeline rather than replacing it.
• Consistent criteria — Every candidate is measured against the same role requirements, making decisions more repeatable and defensible.
• Handover — The solution and its scoping were shared with the client as the basis for progressing to a full build.

With the concept proven, the client has a clear route to move from proof of concept to a production CV triage service, freeing consultants to focus on the candidates most likely to place.