What an AI consultant does, in plain words

An AI consultant works out where your team loses time, and whether AI changes that.

Sometimes the answer is a tool you already pay for, such as Copilot in Outlook, and someone needs to show your team how to use it. Sometimes it is a small tool we build. Often the answer is to leave a task as it is, because the saving doesn't cover what it costs to build and maintain something.

Next to the saving, we work out what building and monthly running would cost, and where your data goes. We don't resell licences and take no commission on tools we recommend.

How the day runs

Nobody has to prepare anything. We look at the work as it happens today.

  1. Beforehand, a 30-minute call. We agree which tasks to look at and who we talk to. Usually three to five people in different roles.
  2. In the morning, an hour with you. Where you think the time goes, what must not go wrong, which tools you pay for and what they cost a year.
  3. Then 45 minutes with each person, at their own desk. They show us how they do an ordinary task. We note the steps and how long each one takes.
  4. At the end of the day, half an hour with you. What we saw, and what we still need to work out for the report.

What the report contains

A document of a few pages. You get it within a week of the day in your office.

Three opportunities
The three places where AI saves the most time, ranked by what they return after costs.
Cost
How many days of work it takes to build, and what it costs each month to run afterwards.
Saving
As a calculation with your own numbers: minutes per task, how often, how many people. You can redo it yourself.
Risk
What can go wrong, who would notice, and which data goes to which service.
What we'd leave alone
What we would deliberately not automate, and why.

Made-up example AI scan report · heating installer, 14 staff · p. 4

Opportunity 2 of 3: sorting quote requests

Now
Requests come in through the web form and by email. Two people read each one, work out what's missing (address, boiler type, photos) and chase the customer for it. Around 100 minutes a day between them.
Proposal
A small tool reads each request, fills in the fixed fields in your CRM and puts it on a list: complete, or missing an address, boiler type or photos. Your team works from that list.
Effort
Indicatively 3 to 4 days of work, then monthly maintenance. AI usage at the provider, estimated at €10 a month.
Saving
2 people × 25 min a day × 220 working days = 183 hours a year, about €8,800 at €48 an hour.
Risk
The tool misreads a request or misses a detail, and the request lands on the wrong list. It never writes to customers. Names, addresses and photos go to an AI model with a European endpoint, under a data processing agreement.
Leave alone
Replies to customers, and anything about price. Your team keeps writing those, because a wrong price or promise costs more than the time a tool would save.
This is what one opportunity in the report looks like. The company and the numbers are made up; the structure is the real one.

After the report, you decide

You can also do nothing after the report. If you want one of the three built, you know the price before we start. Code, accounts and data are then in your name, with documentation another developer can follow.

What the scan includes, and what it doesn't

  • The 30-minute call beforehand
  • A day in your office in Ghent
  • The report, within a week
  • Building is not included. We agree on that separately, if you want it.

The scan has a fixed price. You hear it in the first call.

Outside Ghent, we do it remotely

We hold the same conversations by video, while your people share their screen, and you get the same report.

That works well for email, documents and software. Work on a building site, in a warehouse or at a front desk we would rather see in person.

When to start with something else

  • If you already know what you want built, you don't need a scan. Book a call instead.
  • If your team uses ChatGPT or Copilot without any rules, start with the team training. You leave it with an AI policy too.
  • If you already have an app built with Lovable, Bolt or Cursor, start with a vibe-coded app check.