Artificial intelligence

Not in our projects. The model produces a draft or a suggestion, and an employee approves it and owns the result. The saving shows up as time: the same person handles more cases without rushing.

For most firms it comes down to knowing which AI systems you run, making sure staff have adequate AI literacy and being open with users when they are talking to a machine. If you use AI for recruitment or credit scoring, for example, stricter rules apply. We classify every solution before it is built.

It will now and then - the real question is how often. That is why we turn down tasks where one mistake is expensive and nobody reviews the output. During the pilot we measure accuracy against a minimum threshold agreed in advance. Miss it, and the tool does not go live.

Rarely. Most tasks are solved with an existing model, good instructions and access to your documents, which is faster and cheaper. Custom training is worth it when the subject is so specialised that general models simply do not grasp it.

Far less than a full project - usually two to four weeks on a single task, billed at DKK 895 per hour excl. VAT or at a fixed price. The idea is to test the assumption cheaply before you commit money to licences and integrations.