Supplier invoice OCR: how Next&Go processes 30,000 rental invoices a year

Next&Go brokers vehicle rentals for film productions and sporting events. Around fifteen staff, revenue that has just crossed ten million euros, and between twenty and thirty thousand rental invoices to receive, sort and use every year. Since November 2025 that flow has landed in Koncile, come back classified, extracted and renamed, and routed automatically to the project managers and into the ERP.

There are about fifteen of us, so we are not a large company. What we do have is an extremely high invoice volume: we process between twenty and thirty thousand invoices a year. There is no question of a human handling that by hand. Koncile has revolutionised the subject here, literally.
^{Pierre-Henri Vernay, Deputy Managing Director, Next&Go}

A broker whose raw material is the supplier invoice

Founded in 2015 and based at the Cité du Cinéma in Saint-Denis, Next&Go sits between vehicle rental companies and audiovisual and event productions. It buys the rental from the rental company and resells it to the production, with a dedicated project manager per job, in France and abroad.

That buy-and-resell model gives the supplier invoice an unusual status. It is not an accounting document arriving at the end of the process, it is the document that triggers rebilling to the client and carries the margin. Between fifteen and twenty rental companies feed the flow, Hertz, Europcar, Sixt, Rent A Car, ADA and others, each with its own layout, its own naming and its own line item wording. A single large supplier can account for three to four thousand invoices a year.

What we rest on is receiving the invoices from the rental side. If we cannot take in that flow properly, we do bad work: things get forgotten, things get missed, and clients chase us because we are behind on their jobs.
^{Pierre-Henri Vernay, Deputy Managing Director, Next&Go}

With rapid growth, manual handling stopped holding. One person spent half of every day receiving, sorting and dispatching invoices, work with no added value that did not scale. The tools tested before Koncile required lengthy configuration and did not cover every document type received.

It all starts from a single email address

The setup rests on one simple idea. Next&Go created a single address, shared with every rental partner, to which they send their invoices. That mailbox forwards to the account's Koncile ingestion address, and email attachment extraction takes over with no human handling. The internal rule became a team instruction: for an invoice to be processed automatically, it has to arrive at that address.

From there, every document is classified by type. The account separates rental invoices, traffic and parking fine invoices, damage invoices, contracts, tolls and damage reports, each with its own model. Classification rules are written in plain language, exclusions included: discard documents that are not invoices, discard invoices issued by Next&Go itself, discard any document without an invoice number. Invoice OCR then runs against the matching model, across some forty fields.

Renaming every invoice so it finds its own place

The feature that changed daily work the most was not available at the time of the first demo. Next&Go asked for it at the end of October 2025: the ability to rename a file from its content. It shipped the following week and it now sits at the centre of the setup.

Automatic document renaming rebuilds a normalised filename from the extracted fields, in a fixed order: document type, client reference, number plate, date and time of the event, the supplier's original booking reference, amount. A fine invoice comes out as FPS_GT-339-BD_18-10-2025_12h43_PO7895_35,00€, a contract as CONTRAT_GV-132-BK_08-05-2026_110000885. When a piece of information is absent from the document, the model writes NC rather than picking up an approximate value elsewhere on the page.

That filename then becomes a routing key. A webhook triggers an automation scenario that reads the booking reference and drops the document into the relevant project manager's folder, while invoices with no usable reference land in a review folder. In parallel, the OCR API feeds the company's ERP with the structured data and a link back to the original PDF.

When I prepare payments, I need to find the exact invoice behind each line, and to know straight away what is missing. With filenames built on the number plate and the booking reference, I search by plate and it comes up. Whatever could not be matched lands in a separate folder, and I know that is the only place I need to look.
^{Léandra Saporta, Supplier Payments and Collections Manager, Next&Go}

Breaking amounts down line by line for accounting

The hard part is not the totals. The overall net amount on a rental invoice comes out effortlessly. The value sits one level down, in how that amount breaks apart.

We buy rentals from Hertz, Sixt, Europcar, whoever, and resell them to our clients, and margin rates differ by product. Pure rental days, extra mileage, fuel, one-way drop-off surcharges: all of it has to be allocated correctly in the accounts, with the right VAT rate on each item. That pre-entry work used to be done by hand.
^{Matthieu Constans, Chief Administrative and Financial Officer, Next&Go}

Every rental company names those items its own way, and the wording sometimes varies between branches of the same supplier. So Next&Go's teams write the reading rules themselves, in plain language, field by field, with inclusion and exclusion instructions drawn from the actual wording found on the invoices. They version their models, test changes on their own documents and adjust them without a development cycle or an outside vendor. Onboarding a new rental partner, which happens once or twice a year, is configured the same way.

Half a day freed up daily, and two hires avoided

Roughly a fortnight passed between signature and the first automated flow in production, with no integration project and no heavy technical onboarding. The operational result shows up first as time returned.

The half-day spent every day on receiving, sorting and dispatching has gone, the equivalent of several hundred hours a year redeployed onto work that matters. The thirty to fifty invoices received daily arrive classified and named, and project managers find them in their folder without anyone having opened them. The most significant effect, though, is structural: Next&Go outsources its year-end accounts to an external firm and employs no in-house accountant. Automating data entry and allocation makes that hire less necessary.

With Koncile we can handle the bookkeeping entry and the allocation of our purchases and sales without hiring anyone on the accounting side. Potentially that is two roles saved. It is plug and play: you take a subscription, some credits, and the tool does not even need adapting to the company. You build your own models and off you go.
^{Pierre-Henri Vernay, Deputy Managing Director, Next&Go}

What comes next: accounting integration, matching and checking incoming invoices

Three workstreams are identified for 2027. The first is completing the accounting integration, so the extracted allocation feeds the ledger directly. The second is product code matching, which means loading an item reference table into the platform so every invoice line is mapped to the right account, even when the rental company's wording differs from the reference. That is the block turning faithful extraction into billing control, and it is what makes it possible to measure the gap between what was booked and what is invoiced.

The third is checking the authenticity of incoming documents. A company receiving and paying tens of thousands of invoices a year from around twenty issuers is exposed to familiar risks: the same invoice paid twice, a document from an unreferenced supplier, an amount or a set of bank details altered on an otherwise genuine file. Koncile's document fraud detection answers that risk by analysing the file itself as much as its content, through forensic examination of the PDF metadata and the last modifying software, through visual integrity analysis around an amount or an IBAN, and through business consistency checks. Each anomaly feeds a score with its justifications, returned alongside the extracted fields, so only genuinely suspect documents get opened. At this kind of volume, checking everything by eye is not an option, and a fraudulent invoice once paid is almost never recovered.

Frequently asked questions

You do not need to be an expert, in artificial intelligence or in OCR configuration. Any company that wants to bring in a capable OCR, go for it with your eyes closed.
^{Pierre-Henri Vernay, Deputy Managing Director, Next&Go}

Automate your supplier invoices

If your invoices arrive from dozens of issuers in as many formats and your teams sort, rename and rekey them by hand, send us your real documents. We configure your fields and your reading rules, and you judge on your own invoices.

Book a demo