AI voice agent for customer re-engagement

The problem

A Western Australian meal-delivery service relied on a small Customer Care team to manually remind existing customers to place recurring orders before the cut-off. The volume was difficult to manage manually, and the business needed a respectful way to re-engage an older, opted-in customer base.

The phase-one goal was an AI voice agent that could make the reminder call, confirm the next step or transfer the customer to a human team member. The project had a viability benchmark, but it was a target rather than a confirmed outcome.

What I did

As Digital Producer, I set up the project plan, budget, timeline, invoicing and task allocation in Hive. I was the client’s day-to-day contact for voice selection and helped shortlist Australian female AI voices against the brief.

I coordinated the client, agency team and technical workflow when the order-system API proved more limited than expected. For phase one, the team adopted a semi-manual spreadsheet-based workflow and left direct API integration as a future option.

I contributed to Vapi and n8n configuration in Claude working sessions, helped shape the agent’s conversational prompt and handover material, and supported the ongoing account management after the agent went live. This includes call-list scheduling, usage reconciliation, monthly reporting feedback and regular client reviews.

The result

The live agent is making several hundred outbound calls on operating days. From September to October 2026, daily totals ranged from roughly 170 to more than 700 calls, depending on the call-list size.

A consistent share of calls transfers to the human Customer Care team, typically around 4–8% of calls on a given day. I do not claim order uplift, transferred-call conversion, cost savings or staff-time savings because those figures were not available in the reviewed records.

Tools used

Vapi, ElevenLabs, n8n, RingCentral, Hive and Google Sheets.