Custom AI for home service operators, built around your CRM.

It picks up calls 24/7, asks the same intake questions your best dispatcher would, qualifies the lead against your service area + service catalog, books the appointment directly into Jobber/ServiceTitan/Housecall Pro/Workiz, and texts the caller a confirmation. For the calls it can't handle (complex commercial jobs, edge cases, frustrated existing customers) it escalates to a human within 30 seconds. Most operators we deploy this for see missed-call rates drop from 25-40% to under 5%, with no after-hours staff cost.

Yes — that's table stakes for the build to be useful. We hit the platform API directly: create the customer record (or match an existing one by phone), check real-time tech availability against your dispatch board, hold the slot, and write the job back with all the intake notes. ServiceTitan's API is the most permissive, Jobber's is solid, Housecall Pro and Workiz have a few gaps we work around with a hybrid API + email-parse approach. We'll tell you on the discovery call exactly what your platform supports.

Three escalation paths, configured by you. (1) Live transfer — for inbound calls during business hours, the AI bridges to a human dispatcher mid-call with full context already passed. (2) Callback queue — for after-hours or when staff is unavailable, the AI books a callback slot and texts the customer a confirmed time. (3) Hard handoff — for high-stakes situations (water leak, emergency repair), the AI skips qualification and connects immediately. The "guardrails" are configured during the build, not bolted on later.

For most home-service work, neither applies — you're not handling protected health information or sensitive personal records beyond standard customer contact data. For the cases that do touch compliance regimes (medical-adjacent home care, anything with payment data beyond Stripe/Square tokenization), we run the AI layer through a self-hosted n8n + a private Anthropic deployment with no training opt-in. We document data flow, retention, and BAA paths in the architecture doc before any code ships.

After the initial $5,000 setup + first capability ($5k-$10k), most operators land on a $2,500-$5,000/month retainer for 6-12 months. Inside that retainer: 1-2 new AI capabilities per month, monitoring + alerting on the existing surfaces, prompt + model tuning as Anthropic ships new models, and a standing weekly check-in. Most clients exit the retainer in months 9-12 once the system is stable and their team can handle ongoing tuning in-house.

Three differences. (1) Custom over canned: native AI features in Jobber, HCP, and ServiceTitan are good for the 80% case but useless for whatever's actually unique about your operation — we build for your unique 20%. (2) Multi-step over single-turn: a chatbot answers one question; our agents complete workflows (book the job, confirm with the customer, notify dispatch, log everything to the CRM). (3) Context-aware over stateless: the AI reads your job history, customer notes, and price book as context, so its answers reflect your actual business, not generic advice. The KAOS case study on the Work page is a real example of all three.

Depends on volume and which models. For a typical home-service operator running an AI receptionist + post-call summarizer + a booking agent on Jobber, monthly Anthropic + Voyage embeddings + infrastructure costs run $200-$800/mo at the volumes we see (200-2,000 calls/month). We architect the build with cost ceilings — if a single conversation crosses a threshold, the system caps it and falls back to a cheaper model or a human. The retainer covers the integration cost; the AI inference cost is passed through transparently.