Skip to main content

The AI Assistant — The Primary Interface

For most providers, the AI assistant is Fyndow. Rather than clicking through forms to create a quote, set availability, or chase an unpaid invoice, a provider tells the assistant what they want — "schedule a plumbing job for John next Tuesday at 2pm," "send the invoice for the kitchen renovation" — and the assistant carries it out through the same capabilities the rest of the app uses. Every action is a real operation against a real service, not free-text generation of fake data. This chapter explains what the assistant can do and the confirmation flow that keeps money and messages safe.

The Assistant as the Provider's Co-Pilot​

The assistant is a per-provider agent, scoped to the active business. When a provider opens it, the assistant already knows who they are, which business they are operating, their industry and city, and the current date. It keeps a working memory across conversations, so context carries from one chat to the next and it can remember important facts you tell it.

The assistant speaks plain language and works in plain language. You don't learn a command syntax; you describe the outcome you want, and it figures out which part of the toolkit to use.

What It Can Do​

The assistant reaches across the whole business toolkit. Instead of counting individual commands, it's easier to think in capabilities — the same loop a provider runs by hand, now driven by conversation:

  • Scheduling — review the calendar, set and adjust availability, accept or decline bookings, move appointments, and add notes to a job's timeline.
  • Quoting — draft a quote with line items, send it to a client, and follow it through to acceptance.
  • Invoicing — turn a job into an invoice, send it, and track what's outstanding.
  • Payments and money — check earnings, see what's been paid, and record expenses.
  • Services and products — create and edit what you sell, manage product images, and check stock levels.
  • Orders — review incoming orders and move them through fulfilment.
  • Clients — look up a client's history, find their details, and keep their record current.
  • Messaging — read a conversation and reply to a client.
  • Reviews — see new reviews and respond to them.
  • Communities — catch up on activity in your spaces and post to them.
  • Profile and credentials — update your business details, photos, and verification credentials.
  • Insight — pull a dashboard summary or a morning briefing of what needs attention.

This breadth is why the assistant can run a whole business from a chat box: it can read the schedule, draft and send a quote, convert it to an invoice, check payment status, manage products and stock, respond to a review, post to a community, and remember context for next time.

Reading versus doing​

The single most important distinction is between looking something up and changing something.

  • When you ask the assistant to look something up — your schedule, a client's details, your earnings, your stats — it answers immediately from real data.
  • When you ask it to change something — send a quote, raise an invoice, take a payment, message a client, cancel a booking — it does not act on its own. It prepares the action and asks you to confirm it in the app first.

The assistant is held to this strictly: it never claims something was created, sent, saved, or updated just because it set the action up. The change only counts once you approve it.

The Confirmation Flow​

Any action that moves money, sends a communication, or deletes a record has to be approved by a human. This is what makes it safe to hand an AI the keys to invoices and client messages.

How it protects you:

  1. When the assistant prepares a change, it shows you a clear, human-readable summary of exactly what will happen — "Invoice for $450.00 with 3 line items," "Cancel the booking with John on Tuesday."
  2. Nothing is written until you tap Confirm. Describing an action, or even preparing one, changes no data.
  3. The details of the prepared action are held safely on Fyndow's side, so the amount you approve is the amount that gets carried out — it can't be altered between the moment you see it and the moment it runs.
  4. A prepared action expires if you don't act on it, so a forgotten suggestion can't fire later by surprise.

Actions that require confirmation include creating or sending a quote or invoice, taking a payment, issuing a refund, cancelling a booking or order, messaging a client, responding to a review, posting a comment, and deleting any record. Anything that only reads — viewing the schedule, client details, invoices, stats, or search results — happens immediately, because it changes nothing.

Every action the assistant takes, read or write, is recorded, so there is always a complete history of what it did on your behalf.

What It Cannot Do​

The assistant is powerful but deliberately bounded:

  • It cannot silently change data. Every financial or outbound action waits for an explicit confirmation; describing one in conversation does nothing.
  • It cannot act across business boundaries. It is scoped to your active business, so it cannot read or edit another provider's records.
  • It cannot tamper with an action you've approved. The summary you confirm is the action that runs.
  • It is a provider tool, not a customer-facing chatbot. The manual provider screens still exist as a fallback for when you'd rather tap than type — but for most providers, the assistant is the main way they work.

Why This Design​

Putting an AI in charge of money and messaging is only safe if it can propose but a person must approve, and if the proposal can't be quietly altered between approval and execution. Fyndow's confirmation flow gives exactly that: the assistant has the full reach of the business toolkit, while every consequential action passes through a confirmation gate you control. The result is an interface you can talk to in plain language and still trust with your invoices.