Trades are data, not hard-coded. A shop that does more than one trade gets the combined checklists for all of them.
Every Helper capability wraps something that also has an ordinary manual screen. The manual system is the floor: if the AI is unavailable, your company still operates completely — the schedule, the invoices, the client records and the quote editor are all still right there.
That is the opposite of how most AI products are built, and it's deliberate. The AI is never the only path to anything.
Both run on the same conversation engine. What differs is what each one is allowed to do — a Helper structurally cannot reach a capability outside its own set.
Your office manager and basic bookkeeper. It owns the paperwork — invoicing, payments, client records — and surfaces the things that actually need your attention.
Bookkeeping scope is money in and out, job costs, and billables: good enough that your accountant isn't annoyed. It is not payroll and not tax prep.
Keeps each tech organized and carries data between the field and the office — schedule, parts, job details on the way out; daily reports, hours, and materials on the way back.
Hours are tracked per tech per job, so they're a billing input first and a payroll input when you have employees. A one-person shop is just N = 1 of the same machinery.
Onboarding is a conversation rather than a pick-list — what you do, your service area, what you charge. Your trades are worked out from what you say, and a shop can have several.
Tech availability with owner approval, multi-day visits, and disruption handling when the day changes shape.
Two ways in. Jot a quick onsite quote to finish later, or build the full quote end to end — guided step by step while the quote page fills in live beside you. A draft never advances itself to a finished quote.
Invoices, payments, and what each job actually cost, with GST tracked and categories mapped for tax.
Upload your own documents and your techs can ask against them, with citations — and a refusal rather than a guess when the answer isn't really in there.
A background agent watches the state of the business and surfaces briefings, reminders and nudges: overdue receivables, cost overruns, schedule disruptions, thin margins, GST reminders. It proposes; you confirm.
Bulk import at onboarding, add clients with their history as they come back, or import as you go. Active jobs come across too.
Every action a Helper can take declares its own risk level, and you set the policy against it: do it automatically, ask me first, or never. The check happens before the action runs, not after — and an ambiguous answer is treated as a no.
| Policy | What it means |
|---|---|
| Auto | Looking something up, reading the schedule, pulling a client's history — anything read-only just happens. |
| Ask first | The action is held and shown to you, and only runs when you say yes. A job runs long and the total exceeds the quote: the updated invoice is drafted and waits. |
| Never | Switched off entirely. The Helper cannot take that action on its own, and won't offer to. |
One policy governs both the Helpers and the background agent, so the two can never disagree about what's allowed.
FirstVisitAI is used by you and your techs. Your customers don't talk to it — what reaches them is ordinary documents and messages your shop sends. There is no customer-facing chatbot, and it does not answer your phone.
Monthly subscription, set up by conversation, and your own AI trained on your services, pricing, and availability.
