Customer documents that file themselves into the CRM

Project Overview

Who is LeaseBroker?

LeaseBroker is a Belgian financing broker based in Ghent. The company arranges leasing, renting, and flexible financing for entrepreneurs. Alongside its brokerage business, LeaseBroker runs its own financing company for asset financing and a platform for dealers. It works with several banks and financing partners.

The company is growing fast. But that growth kept hitting the same wall: preparing files was still entirely manual.

The challenge

For every application, staff had to save documents, rename them, check them, and retype the data into the CRM. The documents came in unstructured — by email, through forms, over WhatsApp. ID cards, driving licenses, tax documents, and quotes, all mixed together.

It cost the team a lot of time every day — time that went to admin instead of the client.

"How do we make the prep work around documents disappear, so our people can focus on the client right away?"

Project Delivery

The approach

We started with an A.I. Opportunity Mapping — first an intake conversation, then a workshop with the LeaseBroker team. We mapped the entire workflow phase by phase, including the manual work at each step.

That exercise surfaced six concrete opportunities. Document automation came out as the unanimous top priority — the highest value, and limited complexity.

From there, we continued in an automation retainer. We deliver in two-week sprints, with a review together after each one. That way, LeaseBroker helps decide what comes next.

What we built

1 · Document Intelligence pipeline

One engine that reads, recognizes, checks, and writes incoming documents straight to the CRM. From PDF to CRM in under a minute.

1. The document comes in, either as an email attachment or through a form.
2. The A.I. classifies the document and recognizes its type.
3. The A.I. reads out the fields and rates how confident it is about each one.
4. Business rules validate the output — think checksums, chassis numbers, postal codes, and expiry dates.
5. Fields with high confidence go straight into the right field in the CRM.

The benefit: the prep typing disappears, and a document is processed in seconds instead of minutes.

2 · Review dashboard with a human in the loop

People now only look at what the A.I. isn't sure about. The dashboard shows the document with a box around every value it found, so you can immediately see where a piece of data came from. Fields are color-coded by confidence: green, orange, or red. One click sends the approved fields to the CRM.

The benefit: you stay in control of what lands in your CRM, without having to check everything by hand.

3 · Everything inside the CRM itself

The review sits directly on the file. Uploading and writing data back both happen inside the CRM. No extra tool needed, and nobody has to learn a new system.

The extracted data lands in a separate, protected data layer, so the automation can never overwrite existing data. That was a hard requirement: the human stays the owner of the file.

4 · Automatic naming and filing

Documents are automatically renamed according to LeaseBroker's own conventions and linked to the right file. Hunting for a document is now a thing of the past.

5 · Capturing calls with Leexi

Phone calls with clients are transcribed with Leexi, our A.I. notetaker, and linked to the right file. That's the foundation for the next step: a checklist that fills itself in based on the call.

Technology and security

The solution runs on Microsoft Azure, inside LeaseBroker's own environment. For extraction, we combine Azure AI Document Intelligence with A.I. models we trained ourselves on sample documents per document type. On top of that sits an LLM layer — a language model that interprets free text like contracts and emails.

Because this involves ID documents and financial data, security was a starting point from day one. Data stays within Europe, passwords and keys sit in a secure vault, and every user only sees what they're allowed to see.

We deliberately kept handwritten documents out of scope. The margin for error and the cost don't justify the gain there. We'd rather make that kind of call upfront than after the fact.

Results

The pipeline is live in production. The most common document types are now processed automatically from start to finish, with high accuracy. Processing a document takes seconds instead of minutes. The prep work that used to occupy a team every day has largely disappeared.

Just as important: the foundation is now in place. Adding a new document type takes about half a day to a full day of work. LeaseBroker's entire document flow has been mapped, from the first application to aftersales, so it can be extended step by step.

Flowkify continues to support LeaseBroker as its A.I. partner. The next steps are lined up: automatically turning client calls into a filled-out checklist, smart follow-up on missing documents, and a model that helps predict which financing partner gives a file the best chance.

Want to find out what A.I. could do for your document flow? Get in touch with Flowkify for a tailored A.I. scan.