A working session with David Máj
When tools and pilots multiply but the work does not improve, the problem may not be the model. It may be that the outcome was never made clear enough to build.
Tell me what is stuckWhere the cost grows
It is the time spent trying different directions, checking results, correcting assumptions and rebuilding something that was never clearly defined in the first place.
A company can open an account, try a new tool and produce an impressive result in an afternoon. The difficult part begins when that result has to become repeatable work.
People explore tools and approaches before agreeing what the work should become.
A fast output saves little when someone must repeatedly reconstruct the context and repair it.
A weak definition of the outcome reappears later as changed flows, replaced tools and discarded pilots.
The work before the work
A prompt can improve one run. It cannot decide what the company should build.
A specification is where an initial idea becomes clear enough to examine. What problem are we solving? What should the result look like? How does the work move from one step to another? Which exceptions matter? What must remain a human decision? How will we know that the result is correct?
This is not paperwork before the real work. This is where much of the real thinking happens. The less time we spend defining the outcome, the more time we usually spend correcting the execution.
One decision changes the cost model
AI remains in the work
AI can work with changing context, incomplete information and cases that cannot be described economically through fixed rules alone. The finished service continues to use the model, with limits, approval and a measurable cost per completed task.
AI helps build the tool
AI can help design and build an integration, automation or specialised application. Once it exists, the recurring work may run through predictable code without paying for model reasoning at every step.
Neither approach is automatically better. The work, the variation, the cost of mistakes and the expected result should decide.
One real initiative
Bring one real initiative: a pilot that stalled, a process that still needs too much manual work, an inconsistent tool or an idea nobody has turned into a workable solution.
No AI strategy presentation required.
What did you expect to change, what happens today and where does the work still get stuck?
We clarify the result, the important exceptions, the human decisions and the evidence needed to accept it.
We decide where reasoning adds value, where conventional software is better and what should not be built at all.
You should know what to stop, what to specify, what to test and what is worth taking further.
Before you send it
No. You can bring an existing pilot, a recurring process or an idea you are considering. The useful starting point is one real piece of work, not a finished AI brief.
No. Describe what you expected to change, what happens now and what remains unclear. The consultation starts from the work rather than a presentation.
No. If a rule, integration or conventional application is the more reliable and economical answer, that should be the recommendation. AI earns a place only where it adds useful capability.
David Máj reads it and replies within two business days with the proposed scope, preparation and price for a working session. Nothing is booked automatically.
Start with the situation
David reads every request. You receive a proposed scope, preparation and price within two business days. Nothing is booked automatically.