AI implementation
Implementing AI in your business processes
Implementing AI only pays off when it is attached to a real process. Van de Braak Consultancy builds AI support into existing workflows: reading documents, classifying data, drafting first versions, running checks. Always with an employee who approves the end result.
Where does AI actually save time in a business process?
AI works best on work that occurs often, has a clear input and a checkable outcome. Think of reading invoices and quotes, summarising customer correspondence, drafting first versions, and flagging anomalies in data. Work that requires judgement stays human.
Where does AI not work?
On processes that occur rarely, where the input differs every time, or where a mistake costs money or trust without anyone noticing. Without clean source data it makes little sense either: a model working on messy records produces messy results.
What the EU AI Act asks of you
Since August 2026 the European AI Act requires that people know when they are dealing with an AI system, and that AI-generated content is recognisable as such (article 50). For most business processes that comes down to recording which system does what, and making it visible when a customer is talking to an AI assistant.
When you don't need us
If you mainly want to experiment with a chat interface, existing tools will get you further than a custom project. We only build something once it is attached to a process where time or money demonstrably goes.
How we approach an AI implementation
- 1
Pick one process
We start with one process that occurs often and measurably costs time. That makes it possible to establish afterwards whether the implementation actually delivered anything.
- 2
Baseline measurement
Before anything is built we record how much time the process costs now and how often mistakes occur. Without that measurement, every claim about time saved is a feeling.
- 3
Build with a human in the loop
The first version proposes, an employee approves. Quality stays safeguarded and you see in practice where the model gets it wrong.
- 4
Only scale when it holds up
If the process demonstrably improves, we expand. If it doesn't, we stop. That is cheaper than continuing to build something that doesn't work.
How we price
An AI implementation starts with a bounded first step at a fixed price, including the baseline measurement. Only once that step demonstrably saves time do we agree a follow-up. So you are never locked into a project whose return has yet to appear.
See also what we build and the initiatives we run ourselves.
Discuss your process
Tell us which process costs the most time right now. We'll look together at whether AI changes that.