Giving recruiters back the hours they spent on first-round calls.

The problem

A recruitment consultant's day fills with phone calls that mostly end in a no. The first-round screen exists to check the basics: is the candidate actually available, do they have the right to work, does the salary expectation land anywhere near the brief, are they still interested in the role they applied for three weeks ago.

It's essential work and it's almost entirely scripted. Consultants were spending hours a week on it, calls happened only during office hours, and good candidates went cold while waiting in the queue behind unsuitable ones.

The bottleneck was never judgement. It was the sixty calls you have to make to find the six worth having a conversation about.

What we built

We built a voice agent that conducts the structured part of the screening conversation, and knows when to stop and get a person.

• Natural conversation, not an IVR. Real-time speech recognition and synthesis, low enough latency that candidates talk to it the way they'd talk to a person — including interrupting it.
• Structured screening. The agent works through availability, experience, right to work, compensation expectations and role preferences, adapting the order to how the conversation actually goes.
• Warm transfer. When a candidate qualifies, the agent transfers them to an available recruiter mid-call, with the context already captured. The candidate never has to repeat themselves.
• A recruiter dashboard. Every call recorded, transcribed and attached to the candidate record, so the consultant reviews rather than re-interviews.

How the agents were used

The conversational core, telephony integration and transfer logic is the AIOBC Voice Agent. What we built specifically for this client was the screening logic, the qualification rules, and the dashboard their consultants live in.

The outcome

Screening stopped being a queue. Candidates get called back promptly, including outside office hours, and consultants spend their time on conversations where their judgement actually matters. The transcripts turned out to be an unexpected benefit: for the first time, the agency had a searchable record of what candidates were actually saying about salary and availability across the whole market.

The part we insisted on

The agent screens. It does not decide. Every candidate it declines to transfer is still surfaced to a human, because an AI system that quietly filters people out of a hiring process is a system nobody should deploy. Under current UK rules, automated decision-making in recruitment carries specific obligations around transparency, human review and the right to contest — and we built the human-review gate in from the first sprint rather than bolting it on afterwards.