Internal AI Assistant
Ask your business anything.
Your team asks in plain language and the assistant answers from your own data only — through typed tools, with citations to real records. For any write — a task, a note, a deal move — it asks for your confirmation first, and never exceeds your own permissions.
Grounded Q&A
One plain question — a cited answer
Your staff member types an ordinary question. The assistant calls a tool that reads your real fleet and returns one sentence plus a list of cars, each an openable record. When it proposes a write, a confirm card appears — nothing happens until you tap.
- Typed tools read your actual records — no guessing
- Each car in the answer is a cited, linked record
- Pending writes wait on your explicit confirmation
Grounded in YOUR data
It answers from your records — it never guesses.
A generic chatbot makes things up with confidence. This one can't: every answer comes through a typed tool reading your own data, and every figure carries a citation to a real record id. If it isn't in your system, it says so instead of inventing it.
- Every answer comes through a typed tool that reads your real records
- Citations point to actual record ids — you can open the source
- It declines what isn't in your data instead of inventing an answer
What's revenue this month?
How many cars are on rent now?
What's stuck in Quoted?
The tool palette
Typed tools that read every corner of your business
The assistant doesn't roam your database — it calls specific tools, each reading one part of your system and returning real records.
Fleet status
What's on rent, what's idle, what's in for service — and which vehicles are due for maintenance this week, by plate.
Pickups & returns
Today's handovers, this afternoon's drop-offs and rentals ending tomorrow — so dispatch knows what's coming.
Revenue & metrics
Revenue this month, utilisation, average rental length and how each channel is converting — answered from the ledger, not a guess.
Customer lookup
Pull a customer's history, open balances and past bookings by name or phone — without leaving the chat.
Pipeline stats
Deals by stage, what's stuck in Quoted, and which enquiries went cold — straight from the live board.
Invoices & documents
Find an invoice, check what's unpaid, surface a contract or attach a vehicle document — every record cited by id.
Confirm before it writes
It proposes the action. You pull the trigger.
The assistant can create a task, add an internal note, attach a vehicle document, move a deal and send a customer message — but none of it happens silently. Every write surfaces as a confirm card, and you stay in control.
01Proposes, doesn't execute
It drafts the proposed write — the task text, the deal to move — and shows it for review before anything happens.
02One-tap confirm
Nothing is written to your system until you tap Confirm. Cancel, and no trace is left behind.
03Within your permissions
The assistant can't do what you couldn't do yourself — it inherits the logged-in user's access exactly.
Three authority tiers
You choose how much it can do — per workspace
From read-only, to small safe edits, to the full toolset. Dial it up as you trust it — confirmation and the user's permission limits hold at every tier.
Read-only
The assistant reads and reports. It can look anything up and cite it, but it never changes a single record.
- Look up fleet, deals, invoices
- Summarise revenue & metrics
- Cite the exact records — no writes
Light-write
On top of reading, it can do the low-risk writes — after you confirm each one.
- Create tasks & internal notes
- Attach a vehicle document
- Always confirm-before-write
Full-write
The full toolset — move deals, send customer messages — still gated by confirmation and your own permissions.
- Move deals on the board
- Send a customer message
- Never exceeds your own access
Activity log
Every call, every write
Sara K. queried cars due for service
Sara K. confirmed: create reminder task
Omar F. confirmed: move deal to Booked
Per-user & fully logged
Each person's session is their own — the record stays.
Every user gets a private session that runs strictly within their own permissions. And every query and every confirmed write is written to a full activity log — you always know who asked what and who confirmed which change.
Use our model — or bring your own.
The assistant runs on the platform model out of the box. On Enterprise you can connect your own — OpenAI, Anthropic Claude or Google Gemini — via the API, keeping your data on your own footprint.
See it on your own data
Ask your business a question — get a cited answer back.
We'll connect the assistant to a demo fleet and show you how it answers from real records, cites its sources, and asks for confirmation before it writes — in under twenty minutes.