Introducing AI. Before the tool decides.
Your business is to use AI. The more important question is what for.
Before the first licence, I examine which processes may get faster, which belong cut, and how you will measure whether it pays. That becomes the plan your team and your vendors introduce it by.
ON THE TABLE
Automating processes, introducing AI, saving time.
THE DECISION BEFORE
Which process should stay at all?
WHAT IT COSTS TO LEAVE IT OPEN
A bad process doesn’t get better. It just runs faster, and the licence costs money every month.
“We need to use AI now or we’ll fall behind.”
Those who fall behind are rarely the ones who start later. More often, the ones who make the wrong thing faster.
“A vendor showed us what’s possible.”
Then there is a solution. What’s open is the problem.
“The licences are running. Nobody can say what they bring.”
Then it was bought before anyone settled how it would be measured.
A tool decides nothing. It speeds up whatever it is given.
How I read a business before introducing AI.
Five questions, each answered before anyone talks about tools.
Time. Where does the business tie up hours today that it wouldn’t have to?
Process. Which process carries and may get faster, and which shouldn’t be accelerated at all?
Data. What does the business already know, where does it sit, and what may an outside system see of it?
People. Who works differently afterwards, and who decides whether it got better?
Measure. How will you see after six months whether the spend has paid off?
Three words.
Every process gets one of three words.
Accelerate. Carries today and gets faster with a machine.
Cut. Doesn’t carry. It is abolished before anyone automates it.
Keep. Carries and stays with people, because judgement is needed there.
The list is the deployment plan. It tells your vendors what to build and, above all, what not to.
What a rushed introduction costs.
Licences that keep running because nobody cancels them. Hours in which staff experiment instead of working. Processes that get automated and six months later nobody needs. Data sitting in systems whose terms nobody has read. A team that, after the first failure, no longer trusts any machine.
An AI introduction can work and still bring nothing in.
Before the first licence. And after.
The fit call. Twenty minutes. You tell me what’s planned or already running. Afterwards we know whether a consultation is enough or whether the business needs reading first.
The Consultation. One conversation on the one question: whether, where and in what order. Then I work it through, and at the end comes the Decision Letter.
The Forensic. Time, processes, data, people and measurement read in full, every process with one of the three words. Afterwards: the Report, handed over bound. It is the deployment plan for team and vendors.
The Report says whether it stays at that. One thing can follow. It doesn’t have to.
The Direction Mandate. Once it’s being introduced: I hold the direction while team and vendors build, and every quarter I check whether each licence pays. Whatever doesn’t gets cancelled.
What I don’t do.
I don’t sell tools, train teams or program applications. I make sure that before the purchase it is settled what for. Your vendors and your team work to it, and better, because they know what won’t be built.
One case.
AI bought before anyone had decided what for.
Case 03 · €25,000 for a process that no longer existed six months later
Came with: “We need to use AI now or we’ll fall behind.”
Underneath: A vendor was faster than the decision. What got automated was what happened to be running, not what was meant to carry.
Decision: First the foundation examined, then talk of tools. The automated process no longer belonged. Cut without replacement, tool included.
Effect: One process less, one tool less, no running costs for it. Since then the rule in the business: first the use case, then the tool.
Further reading: AI accelerates, the mistake included.
Before your business uses AI, it is settled what for. And what goes in return.
Twenty minutes. Afterwards you know whether a consultation is enough.