Perspective
What AI consulting delivers - and where it stops
AI consulting is a broad word. It covers everything from a keynote to a team sitting in your building for six months. Anyone buying it should know in advance which of the three kinds they are getting - because only one of them leaves a running routine behind.
Last updated: August 2026
01
Three things that all go by the same name
- Orientation
- Talks, workshops, training. They answer what is possible today and raise understanding in the business. They are valuable where nobody has a picture yet - and they leave nothing behind that works.
- Strategy
- Potential analyses, roadmaps, prioritisation. The result is a document. Whether it can be implemented hangs on a question the document rarely answers: can anyone reach the necessary data.
- Implementation
- Somebody builds the routine, connects it to the existing systems and runs it afterwards. Only here does anything change on a Tuesday morning.
All three are legitimate, and none replaces the others. The expensive mistake happens when a business buys the second kind and expects the third - it then has a good document, a list of prioritised processes, and nobody to build them.
The test is simple: what stands at the end of the project - a document, or a routine that runs without supervision?
02
Why the gap hurts most in mid-sized companies
A corporation buys strategy and implements it with its own IT - the department is there, it has developers, and the road from paper to system is internal. In a mid-sized company that road is the bottleneck. IT often consists of one to three people keeping operations running; capacity for a development project is not there, and it cannot be hired for a single undertaking either.
Which is why most AI plans in mid-sized companies stall neither on the technology nor on the will, but between insight and implementation. The finding sits in the document, and then nothing happens for a year.
03
Two gaps, not one
The second gap is named less often and is the more valuable one: nobody knows WHAT should be done either. People inside the business know their own routines well - but not which of them a machine could carry today. That is not a gap in knowledge about the business but about the state of the technology, and it can hardly be closed from inside.
Anyone who sees both gaps also sees why the order is rarely advice-then-implementation. What a routine demands technically co-determines whether it pays; whoever does not know the build prioritises by instinct.
04
How to recognise consulting worth having
- It tests data access instead of assuming it
- A real export, a real file, a real interface call. A promise in the vendor's data sheet is not access, and the difference can cost the entire project.
- It also says no
- If nothing is worth doing, that is what the result says. Consulting that always leads to a project is not consulting but a sales appointment.
- It names limits that are not the model
- Too little repetition, too little volume, unsettled responsibilities. Whoever names as a limit what models supposedly cannot do is rarely up to date.
- It budgets for running costs
- Models change, interfaces disappear, processes move. A plan without an operations line has not been costed to the end.
Here that stage is called the audit and costs €2,500, credited in full against the implementation. That credit is not a discount but the removal of a conflict of interest: whoever earns from the audit has an interest in many audits. Whoever credits it has an interest in usable results.
Frequently asked
What does AI consulting cost for a mid-sized company?
For a dependable assessment with a data access check and a roadmap, budget a mid four-figure sum; with us it is €2,500, credited against the implementation. Talks and workshops sit below that, strategy projects from large firms considerably above.
Do we need consulting first or implementation straight away?
Both in one. Awarding them separately means the effort estimate comes from someone who does not have to answer for it - and that the first real test of data access happens inside the implementation project, which is where it is most expensive.
What sets you apart from a classic management consultancy?
We build what we recommend and run it afterwards. That lets us test whether we can reach the data before quoting - and it is why we advise against a process that will not pay. We give no talks and run no courses.
Do you also train our employees?
Instruction on the systems we build is included. General AI training we do not offer - that is a trade of its own, and there are people who do it better.
Where it gets concrete
Use cases on this topic
Tell us where your time goes.
In the audit we record the process, work out what is worth automating, and you get a plan with fixed prices - credited against the build if you go ahead.