Call your own service line Saturday evening. Voicemail? The AI answers in one ring, asks the questions your dispatcher would — no heat, or no cool? — books the window and texts the confirmation.

Model number, symptom, access notes, gate codes: collected in the first conversation, attached to the work order — the tech arrives knowing what he’s walking into.

Day-before confirmations and on-the-way notices in the customer’s language: fewer locked doors, fewer wasted truck rolls, a route that holds together.

The tech’s voice notes and photos from the job become a clean service report in your template: work performed, parts used, recommendations — drafted before the truck leaves the lot.

The AI drafts the claim from the job record, cites the service bulletin that covers the failure, attaches the photos — fewer rejected claims, fewer Friday afternoons lost to resubmission.

Every AI-drafted document — the report, the claim, the callback analysis — lands in Output awaiting review. Your service manager signs; the system remembers.

Scanned manuals from 1998, PDF bulletins, photos of wiring diagrams, the OEM’s spec sheets — the ingestion pipeline turns the filing cabinet into queryable knowledge.

When bulletin SB-2024-11 supersedes the manual’s procedure, priority order makes sure the answer cites the bulletin — not the page it replaced.

The apprentice sees the procedures; the service manager sees the warranty terms and margins: answers scoped to role, enforced server-side — and OEM NDA boundaries hold.

Warranty claim, callback review, bulletin digest, maintenance-due report: work types molded to your operation, with default libraries and examples per type.

The AI’s workload is managed like the crew’s: one-click views show what is urgent, late, today and tomorrow — nothing slips while the trucks are out.

Every work is a full session: reasoning streamed, job records cited, milestones tracked, linked to the customer and the unit it belongs to.

Lockout-tagout, refrigerant handling, commissioning steps, the ladder-safety policy, the callback-prevention checklist: one library, current revisions on top, superseded versions outranked.

“What is the startup sequence after a compressor swap on this model?” — answered in seconds with the checklist attached, at the unit, on a phone, gloves half off.

Work types carry their default procedures: the AI drafting a commissioning report reads from the current commissioning checklist — by construction, not by memory.

New techs learn your job workflow, documentation standards, safety procedures and the systems on the truck through guided training — productive in days, not a season.

“How do I document a refrigerant recovery?” — asked to the assistant from the jobsite, answered from the procedure, without a radio call to a busy senior tech.

Private chats work on any phone with a login — the how-to knowledge rides along to the rooftop, the basement and the highway rest stop in between.

Job closed → the AI reads the tech’s notes and photos, drafts the service report in your template and the plain-language customer summary, and queues both for review.

New OEM bulletin arrives by email → filed to the right library, affected models identified against your install base, a digest drafted for Monday’s toolbox talk.

Day-before → confirmation in the customer’s language; morning-of → the on-the-way notice. Rescheduling requests flow back to the calendar, not to a sticky note.

Units past the service interval, warranties expiring in 60 days, maintenance contracts up for renewal: audiences built from your own equipment and job history — no list ever exported.

The tune-up offer reads naturally in Spanish to the restaurant owner and in English to the property manager — your shop’s voice, their language.

Opt-outs and contact preferences are enforced by the system, not by memory. Every send is on the record.

Work orders, customers and equipment mapped field by field, with a dry-run preview before anything writes — push, pull or two-way, your choice per object.

Salesforce and HubSpot connectors sync commercial accounts and maintenance-contract leads push, pull or two-way — sales and service finally read one story.

A full API for whatever else the shop runs: the parts supplier portal, the accounting export, the GPS feed — data in, results out, under scoped keys.

The server redacts what a role cannot see — the subcontractor sees the checklist for the job at hand, not the OEM manual library or the customer’s contract terms.

A searchable log of every audited action: who queried which manual, who exported the customer list, who changed a procedure, and when — the NDA audit answers itself.

Five-level workgroups map branches, crews and OEM programs; the air-gap-capable on-premises option puts the manuals physically inside your building, updated by USB.

Ask the chat: “my AC quit — can someone come Thursday?” — watch it qualify the problem, offer windows and confirm like your best dispatcher.

Ask: “how would a tech get a torque spec from the field?” — it walks the cited-answer workflow it would run in your tenant, manual page and all.

The full demo loads a field-service tenant with sample manuals and jobs: spec queries, service reports, warranty drafts — your fleet, working.