Every Unbooked Tenant Called Twice A Day. Every Slot Checked Against The Route. Zero Phone Tag.

Arc Group is a UK electrical and plumbing contractor working the rental-property repair chain. Tenants report faults to their property managers and letting agents; the agents pass each job to Arc Group to schedule and complete. New jobs land in their field service platform every day, from routine repairs to compliance work like electrical installation condition reports. The business runs on schedule density. Revenue is a function of how many jobs each qualified tradesperson completes in a day, and that depends on two things the office cannot fully control: drive time between properties, and whether a tenant picks up the phone.

Scheduling a repair sounds simple until you list what a valid booking actually requires. The tradesperson has to hold the right qualification for the job type: a compliance inspection needs an electrician or a tester, not whoever is free. The visit has to fit the day's route, with real drive time between the property before it and the property after it. A tenant who will not be home can only take a slot late enough for the keys to be collected first. And none of it matters until a tenant answers the phone. The office carried that whole puzzle by hand: call the tenant, miss them, get called back at the wrong moment, re-check the diary, start again. Hiring another scheduler adds another person holding the same puzzle. The constraint was the puzzle, not the headcount.

• Every booking had to clear four checks at once: trade qualification, route feasibility, tenant presence rules, and tenant availability
• Reaching tenants meant repeated manual call attempts around the office's other work
• Compliance job types could only be assigned to specific qualifications, checked by hand
• Unbooked jobs sat in the system while phone tag played out, every one a tradesperson-hour unsold

An outbound AI voice agent that works the unbooked-job list twice a day and books tenants straight into the schedule. Jobs flow from the field service platform into the pipeline automatically. Twice a day the system pulls every job with no booking and places a batch of calls. On each call the agent confirms the property, describes the work and how long it takes, asks whether the tenant wants to be present, offers a morning or afternoon window, and presents up to five concrete slots, each one already survived the full puzzle. The route check hit a hard engineering wall: the drive-time API caps each request at 25 comparison points, so the fix was a two-stage filter, straight-line radius math to discard the obviously infeasible options first, with true drive time computed only on the survivors.

• Twice-daily automated call runs against every unbooked job, batched, with tenants who book dropping out of the list
• Qualified-trade matching: compliance jobs route only to electricians and testers, by construction
• Two-stage route check: radius pre-filter, then real drive time between the surrounding visits
• Presence rules enforced automatically: tenants who will not be home only see slots after key collection
• Graceful failure: any mid-call error hands off to SMS with a self-booking link instead of a dropped call