On any customer, AI suggests the next 1-3 services to pitch.

Deep dive · 01

The model ranks; the catalog keeps it honest

When you click Generate, a server query assembles the real picture for that customer — their most recent vehicle on file and their last several appointments with service names, categories, dates, and statuses — alongside your active service catalog with prices. That bundle goes to the fast AI model, which selects and ranks one to three services to pitch next and writes each one's reason and a short SMS pitch. The selection is the model's; the code's job is to constrain it. Any suggestion whose service id isn't actually in your catalog is dropped before it reaches you, and the list is hard-capped at three. So the recommendation is genuinely AI-ranked, but it can never point at a service you don't offer.

• Reads the customer's vehicle + recent appointment history and your active catalog
• The model chooses and ranks the 1-3 picks — selection is not a fixed rule
• Results filtered to real catalog service ids; invented ids are discarded
• Hard-capped at three suggestions, each with a reason and an SMS pitch line

Send SMS or Build quote

Each card has two actions — Send SMS opens the composer prefilled with the pitch, Build quote opens a new quote prefilled with the customer and service.

You send; nothing is automatic.