AI mustn’t make mistakes with amounts.

Correct, and that’s why amounts don’t come from a language model but from text recognition and structured e-invoices. The AI proposes, a person approves, and everything above a threshold gets a second pair of eyes. A residual risk remains: that’s why we keep spot checks and logs, so you can trace every entry.

We’ve just switched to a new accounting package — do we have to replace it again?

No. Exact, AFAS, Moneybird and Twinfield have good integrations; we build in between and around them. Old custom-built software without an integration is harder: then we tell you honestly up front what is and isn’t possible.

This costs money. Does it actually pay off?

The invoice flow is the most measurable part of your business: minutes per invoice, days to payment, number of reminders. We agree up front what we measure and start small. If it delivers nothing after the first cycle, we stop. Not every project pays off; we say that too.

Our admin is different from everyone else’s.

Your ledger and your approval route are yours; the steps are the same everywhere: invoice in, check, book, pay, follow up. The AI learns from your history, not from a standard. What is truly unique stays human work.

Will our administrative assistant become redundant?

The retyping disappears, the work doesn’t. Vacancies for simple administrative work have been declining for years, while demand for people who check, solve exceptions and talk to customers is growing. Whoever retypes now will do that work later. That’s a conversation you have with your team, and we help with it.

Feature · Administration & finance

A quarter of all invoices are paid late. And your admin team only sees it once it’s too late.

Administration is the department with the most manual work nobody sees. Invoices from the mailbox into the accounting package, tying bank lines to invoices, expense claims from WhatsApp, and once a month the VAT scare. The work isn’t hard. There’s a lot of it, it’s dull, and one mistake costs more than a day of typing. Exactly the kind of work you hand to a machine, with a person checking.

// A report from a fictional building services contractor with 28 employees, one administrative assistant and a managing director who checks the invoices in the evening. The company is fictional, the situations are real: we run into them again and again.

Book an intro call Take the free AI scan

Sound familiar?

Friday afternoon: the pile of PDF purchase invoices from the mailbox is retyped by hand into the accounting package.
Sales invoices go out as PDFs, without a payment link. Paying is a chore for the customer.
Nobody looks at the debtors list weekly; the reminder only goes out when the money is really needed.
Monday morning: matching bank transactions to invoices one by one.
The weekend before the VAT return, the hunt for receipts begins.
Expense claims arrive as photos in WhatsApp and get retyped later.
“Our bank account number has changed” by email, and the master data is updated without calling back.
The same customer address lives in three places: the CRM, the accounts and the invoice. And one of them is wrong.

What usually runs here

The software we come across here most often. It does not have to go; we build around it and in between.

Exact Online accounting, bank, debtorsAFAS ERP and financeTwinfield accounting via the accountantMoneybird invoicing and bookkeepinge-Boekhouden.nl online bookkeepingSnelStart SME accountingBasecone scan-and-recognise purchase invoicesYuki accountancy platformPeppol e-invoicesMollie payment links and direct debitBank integrations Rabo, ING, ABN, bunqSimplicate / Teamleader hours, projects, invoicingExcel / Google Sheets timesheets, cash flow

// the same working day, twice

A day in administration & finance: now and next

On the left, the day as it usually runs now. On the right, the same day once the dull parts are automated and AI does the groundwork. Watch what changes: not the people, but where their time goes.

// how it goes nowWithout AI

08:15AdministrationMailbox open. Fourteen purchase invoices as PDFs, three in a zip, one as a photo of a receipt. Print first, then retype.
09:40AdministrationYesterday’s bank transactions: 31 lines, looking up one by one which invoice they belong to. Two are left over: “unknown”.
11:00Managing directorCalls from the building site: “Has Van der Berg paid yet?” Nobody knows for sure; the list is from last week.
13:30AdministrationEmail from a supplier: new bank account number. Gets updated in the master data. No phone call.
15:00InstallerSends a photo of a fuel receipt via WhatsApp. It won’t make it into the expense claims until next week.
16:30AdministrationCreating this week’s sales invoices: copying customer details from the CRM, emailing a PDF, no payment link.
19:30Managing directorAt home, laptop open: checking invoices and going through the debtors list one more time. Reminders tomorrow. Or the day after.

// how it will goWith AI and automation

08:15AdministrationThe fourteen invoices were already read overnight: amounts, VAT, ledger account and purchase order matched. Eleven are ready with “approve?”, three need a look because the amount doesn’t match.
09:40Administration29 of the 31 bank lines were matched automatically. Two sit in the queue with a suggestion. Five minutes instead of an hour.
11:00Managing directorChecks his phone: this morning’s debtors overview. Van der Berg paid yesterday; two others automatically received a friendly reminder.
13:30AdministrationThe IBAN change is held until someone has called the supplier’s known number. The AI puts a red label on it: unusual pattern.
15:00InstallerPhotographs the fuel receipt in the app. Amount, date and VAT are recognised; the expense claim is ready for approval straight away.
16:30AdministrationSales invoices go out as e-invoices, payment link included, with customer details from a single source. No retyping.
19:30Managing directorLaptop closed. The weekly overview landed in his inbox at 17:00: outstanding, paid, what needs attention.

Where AI makes the difference here

Per use case: what the automation does, what a person keeps doing, and whether it is proven or still promising. We say which one honestly.

01proven

Reading and pre-posting purchase invoices

AI reads the PDF or e-invoice, recognises amounts and VAT, picks the ledger account based on your own history and matches it with the purchase order.

Still human: approves every proposed entry; deviations above a threshold get a second pair of eyes.

Working more efficiently Saving time

02proven

Matching bank transactions automatically

Transactions are linked to invoices by amount, reference and customer; whatever doesn’t fit lands in a queue with a suggestion.

Still human: works through the queue and decides when in doubt.

Saving time

03proven

Debtors that chase themselves

A payment link on every invoice, automatic reminders on fixed days, and a prediction of who will pay late based on payment behaviour.

Still human: decides on a reminder, payment plan or debt collection, and calls the customers that matter.

Saving time Handling more with fewer people

04proven

Fraud lock on master data

A changed bank account number, or an invoice that deviates in amount, frequency or layout, is held and flagged. The rules and the workflow are proven; smart anomaly detection is the next step.

Still human: calls back on the known number and releases it. Technology enforces the four-eyes principle, people do the checking.

Better security

05proven

Expense claims and hours without retyping

A photo of the receipt becomes an expense claim with amount, date and VAT. Timesheet suggestions from calendar and planning are still promising, not standard.

Still human: confirms and corrects.

Saving time

06proven

Year-end without the shoebox

Everything the accountant needs is ready and structured: entries with their source, approvals, contracts. AI writes the explanation per item.

Still human: the accountant remains ultimately responsible and reviews; you no longer hand over a box in March.

Working more efficiently

25%

of Dutch invoices were paid late in 2025; only Ireland did worse. [1]

14 days

is how long an incoming invoice in the Netherlands takes to process on average. [2]

44%

of businesses in the Netherlands and Belgium have faced invoice fraud. [3]

Honestly: what AI does not do here

AI doesn’t read minds: if your chart of accounts is a mess, the AI learns the mess. Tidy up first, then automate.
A recognition rate of 98% sounds good, but it means 2 in 100 invoices go wrong if nobody looks. That’s why a person approves every proposal and spot checks remain.
You don’t stop fraud with software alone. The call-back rule for a new bank account number is the real security; we make sure you can’t get around it.
The accountant remains ultimately responsible. What changes is the question they ask: not “is this right?” but “how did this figure come about?” And now you have an answer to that.

The objections we hear — and whether we can close them off

We looked them up and heard them from clients. For each objection we say whether it is really solvable, partly solvable, or a risk that stays and that you accept knowingly.

Can it be closed off: yes. We work in a business-grade AI environment in the EU, under a data processing agreement and without your data being used to train models. Over the past two years, the Dutch Data Protection Authority received dozens of data breach reports caused by employees pasting data into free chatbots on their own initiative; that is exactly what a dedicated environment and clear agreements prevent.

// how we start, the first thirty days

01Week 1: core session in admin. We follow the invoice flow from mailbox to payment and count the actions.
02Week 1: fraud lock and call-back rule on master data. Costs almost nothing, prevents the biggest damage.
03Weeks 2 and 3: reading and pre-posting purchase invoices automatically in your existing package, with approval by your employee.
04Weeks 3 and 4: payment links and automatic reminders on sales invoices; bank matching switched on.
05Week 4: measure. Minutes per invoice, days to payment, number of manual entries. Only then the next step.

// what it costs and how it works

First the free AI scan, then a core session on your floor (a half-day of 4 hours, €596 excl. VAT and travel costs) with a core report and an honest go/no-go. We build on a project basis with fixed hours: €110 per hour for straightforward work, €165 for complex development.

More hours only with your written approval. If we cannot deliver, you pay nothing for what was not delivered.

AI scan & strategy Process automation Custom software & integrations

Want to know what this means for your administration & finance?

Take the free AI scan or book an intro call with someone who builds it themselves. A reply within one working day, no slides, no obligations.