How Jeffway built an AI-native operation for tenant booking

On the same repairs desk that now handles more than 250 jobs a month, actually booking the tenant in was a slow job of its own. Coordinators made the calls in the gaps between everything else, tenants often weren't in to answer, and appointments quietly slipped by a day or two.

The job itself was already in the system. The one piece missing was a time the tenant had agreed to, and getting that agreed could take several attempts and a lot of voicemail.

For the jobs a coordinator has marked as ready, a voice agent now makes the booking call. The coordinator stays in charge of what goes out. Nothing is dialled unless they've said the job is ready, but they no longer have to sit and make the calls themselves.

The agent talks to the tenant, offers a time that suits them, and writes the agreed appointment straight back onto the job while still on the phone. Nobody has to type it up afterwards, and the booking is on the record the moment the call ends.

If a tenant doesn't pick up, the agent simply tries again later, so a missed call no longer depends on someone remembering to ring back. And if a tenant would rather not be contacted this way, that's respected.

That leaves the coordinators to do the part that actually needs judgement: deciding when a job is ready and dealing with anything out of the ordinary, instead of spending the afternoon redialling the same numbers. The appointments that used to slip now tend to get made and recorded the first time round.