A vendor sends a quote for the roof repair on unit 12. It arrives the way these things always arrive — a PDF attached to a WhatsApp message, sometimes a photo of a printout, occasionally a forwarded email. Someone on your team opens it, reads the line items, checks the subtotal, recalculates the IVA, finds the right work order, and types the numbers in.
That last step is the one worth looking at. The data is already in the document. The vendor wrote it down. Your team is re-typing information that already exists, and every re-typed figure is a chance to transpose a digit.
Why vendor coordination eats the morning
Coordinating a single maintenance job touches more channels than it should. The photo of the problem comes in on WhatsApp. The quote comes back as a PDF. The follow-up is a phone call. The approval is a reply somewhere in the thread. By the time the work is awarded, the trail of who-said-what is scattered across three apps and nobody's sure which version of the price is current.
The chaos isn't the vendors' fault, and it isn't WhatsApp's fault either — WhatsApp is where your vendors actually are. The problem is that none of it lands anywhere structured. A quote in a chat thread isn't a record. It's a thing you have to find again later.
Systemise the coordination first
Before any of the clever parts, the coordination itself has to live in one place. In BasePro, when an inspection finds an issue, the maintenance order is created against that specific asset, and the vendor is notified through a structured, pre-approved template — by email or WhatsApp Business, whichever channel that vendor uses. The operator triggers it; nothing goes out without that. WhatsApp is one delivery channel among several, not the system of record.
That alone turns a scattered thread into a tracked job: one order, one vendor, one place where the quote and the confirmation belong.
Where the AI actually sits
Then the quote comes back, and this is the part that genuinely changed.
Upload the vendor's PDF. BasePro reads it — the line items, the totals, the IVA, the payment terms — and surfaces a clean, structured summary against the work order. Under the hood that's Gemini 2.5 Pro doing the extraction; what you see is the numbers already pulled out, ready for review. You check them against the job and confirm. The extraction is automatic. The approval is yours, every time.
This is what AI is good at: reading an unstructured document and turning it into structured data. It is not deciding to pay anyone, and it is not picking the vendor — the vendor ranking in BasePro is plain rule-based logic, not a model, and the maintenance request form is a form. The intelligence is narrow and honest: it reads the quote so your team doesn't have to re-key it. PDF Quote Intelligence ships today on the Enterprise tier.
The outcome is the point, not the model. The quote that used to take a careful few minutes to enter — and occasionally a fixed typo a week later — becomes a glance and a confirmation.
The honest horizon
There's a version of this that goes further: a drafting layer that surfaces vendor follow-ups, compares competing quotes, and summarises a thread for your review before you reply. We're designing it. (In design — on the roadmap, not shipped.) When it arrives it will follow the same rule the rest of the platform follows: it proposes, you approve. Nothing reaches a vendor, a payment, or a stakeholder without an operator signing off.
We'd rather tell you what's shipped and what's on the way than blur the line between them.
Where BasePro fits
Vendor coordination, quote capture, and the work order it all attaches to run on one platform, so the quote you confirm today is already in the ledger when you close the month. The re-keying step doesn't get faster — it goes away — and the audit trail is intact because the record was structured from the moment the PDF was read.