A service operation is a stream of small timing decisions — which job first, which contract next, which breach is coming — made all day by tired people. The doctrine here: deterministic signals, AI delivery, human approval. Engines make the schedulable decisions on schedule; agents surface the rest with evidence; dispatchers stay in charge.
What runs before the shift starts
- Preventive jobs generate themselves — schedules and contract campaigns create the work the contract promised, ahead of due.
- Renewal leads raise themselves — expiring contracts enter the horizon with service history attached.
- Optimize day fills the Gantt — the deterministic scheduler packs free slots from the backlog, respecting skills and windows.
- Nudges publish overnight — jobs drifting toward SLA breach, contracts consumed past entitlement, unbilled completed work, the quote nobody followed up — queued with evidence by morning.
What agents surface — with the working shown
Sense answers the questions a service head asks at month-end, any day: "which contracts breach SLA most often?", "which technicians' jobs bounce back?", "what's unbilled and why?" — grounded in the operation's own records, scoped to the asker's permissions. It drafts customer follow-ups and renewal outreach; writes wait for approval. The daily brief opens each dispatcher's day with their queue's facts, ranked.
Agents you configure, not commission
AI Studio builds service-specific agents on your own knowledge — fault-tree assistants for technicians, triage helpers for the desk — with budgets enforced before model calls, bounded tool access, and full audit. Scheduled agent workflows can watch the contract book and publish suggestions autonomously; dispatch itself stays deterministic tooling, and acting on any suggestion is a human click. Governance: AI trust.
Common questions
Does AI dispatch the work?
No — dispatch is deterministic tooling: skills, presence, capacity and load, inspectable and overridable. Agents surface patterns and draft suggestions around the schedule; the dispatcher decides. That division is deliberate and documented.
What service work generates itself?
Preventive jobs from schedules, campaign sweeps across the contract base, renewal leads at the horizon, claim assembly from completed jobs, and scheduled nudge sweeps over breaches, entitlement and unbilled work.
How would an agent help with SLA management?
By reading milestones and job history it surfaces the pattern — three remote-route AMCs driving most breaches — and drafts the mitigation (window renegotiation, coverage change) for a human to approve, before the quarter's numbers land.
Can technicians get AI help in the field?
Yes — assistively: agents grounded in your manuals and fault trees answer diagnostic questions on the device, supporting the technician's judgement rather than replacing the procedure.