What an AI and process scan delivers

Why measure first

Most businesses know that their administration is too heavy. A February 2025 Visma study found that 41% of Dutch business owners spend more than half their week on it. Almost nobody knows exactly where those hours go. Ask three employees and you get three answers.

Without measuring, you start on gut feeling. You build something for the process that stands out most, not the one that costs the most hours. And you have no baseline, so you never find out whether it worked.

Our rule: no build before the number. Count first, then rank, then choose.

Week 1: map the work with your people

The scan starts with the people who do the work, not with the systems.

1. Kickoff, half a day, on site. With the owner or operations manager and three to five key employees: sales support, administration, planning. We put every recurring process on a whiteboard: what comes in, what goes out, which systems sit in between, who does what.
2. Shadowing. In the days after, we sit with the people who run the processes. Not to judge, but to see what really happens: the extra spreadsheet nobody mentioned, the phone call every order costs, the waiting for a colleague.
3. System review. Which packages run (Exact, AFAS, Twinfield, a TMS, a WMS, custom software), which have an integration, and where data lives that exists nowhere else.

By the end of week 1 we have a process overview: usually 30 to 60 processes for a business with 10 to 100 employees, each marked manual, partly automated, or automated.

Week 2: count hours and rank bottlenecks

Now it gets concrete.

• Hours per process. For every manual process we count, together with the person who does it: how often per week, how long each time, how many people. That gives hours per week per process.
• Errors and waiting time. Where it goes wrong, how often, and what an error costs (a return, a disputed invoice, an angry customer). How long something waits before someone picks it up.
• Feasibility. How structured the source is, whether the target system has an integration, how many exceptions there are.
• Risk. Whether personal data is involved, whether it touches a decision about people, and which GDPR and AI Act conditions apply.

Each opportunity gets a score on hours × feasibility × risk. The ranking that comes out does not put the most exciting process on top, but the one that returns the most the fastest.

Week 3: roadmap and first prototype candidate

In the last week we write up the result and present it in one hour to management and the employees involved.

• The roadmap. Which opportunities to take on in the next 3, 6, and 12 months, in which order, and why.
• The first candidate. One process that can become a working prototype in 10 business days, with an estimate of hours before and after.
• Conditions. Per opportunity: which data, which integration, where a person keeps approving, and what has to be arranged for GDPR.
• The decision. Prototype, workshop, or stop. All three are a good outcome.

What the result looks like

The most important part of the result fits on one screen. This is an illustrative example for a wholesaler with 45 employees:

Below that comes the top five, with the hours, feasibility, risk, and proposed approach for each opportunity. As a diagram it is a grid of 47 blocks: 14 marked as an opportunity, 6 of those with an AI step, and 3 ringed as a quick win. That is the picture you take back to your management team. Not “AI is interesting,” but “430 hours a month sit here, and this one goes first.”

What we need from you

A scan is not a project that brings your team to a standstill. Concretely:

• A half-day kickoff with three to five people.
• Two to three hours of shadowing per key employee, during their normal work.
• Access to a test environment or read access to the systems, arranged through IT.
• One hour for the presentation.

In total about one and a half days of your employees’ time, spread over two to three weeks. And one thing that costs no time but matters: tell your people the scan is about the retyping, not about them. They know better than anyone where the time leaks away, and they are glad to help once they know that.

When you do not need a scan

A scan is the normal route, but not always the best one.

• You already have one clear process that obviously costs the most hours, with a structured source and a system that has an integration. Then you can have a prototype built right away; the scan would only confirm what you already know.
• You have fewer than 15 employees and three processes in total. Then a 30-minute intake call is probably enough.
• You want to bring your team along first. Then a workshop is a better first step; the use-case list that comes out of it is a light version of the scan.

Not sure? That is exactly what the free intake is for.

Example

A technical wholesaler with 45 employees had a scan done. Order entry came out on top: 42 hours a week, 2.6% errors, and a structured source in the mailbox. The prototype that followed brought that down to 9 hours a week. An accounting firm with 18 employees found three processes that all hung on the same Excel tracker, and tackled them in one go. Figures are illustrative. Read the cases order entry from email and document collection.