70 per cent machine. 30 per cent human.

Two days in the MBA in Technology & Innovation at the University of Liechtenstein. The topic: business models and customer experience. Asked how they work on business models today, the participants answered with a figure: more than 70 per cent of what is produced in business model development now comes from the machine.

The figure measures output. Not impact.

Whoever develops something also bears responsibility for it. For a long time, these two aspects of professional work were closely intertwined.

For a long time, the two were closely linked. Anyone who worked on a model for weeks, questioned assumptions and traced calculations step by step automatically built up a deep understanding of its strengths, limits and risks. The development process itself trained judgement – responsibility arose, so to speak, from intensive engagement with one’s own work.

This link has been broken. Today you can have a complete business model in front of you in twenty minutes and three days later still not know whether any of it is right.

The remaining 30 per cent are therefore not a remainder. They are the thing itself. Which of the proposed patterns fits the capabilities of this company rather than just any company? On which single assumption does the entire model rest? These are questions of judgement and of contact with reality — a price list, a conversation, a calculation of your own. The machine can ask them. It cannot answer them.

For work on business models, this means the burden of proof has shifted. In the past, you had to show that you could put a model together. Today, you have to show that the one in front of you holds up. That is the more demanding task, and it cannot be delegated, because it requires someone who knows the company and is prepared to be wrong.

Anyone who has the 70 per cent but not the 30 is holding a draft. Not a business model.

Testing a model for robustness

If the burden of proof has shifted, you need a method for providing that proof. A good starting point is to ask for the one assumption on which the whole model rests. Almost every business model has such a load-bearing assumption – about customers’ willingness to pay, the cost of a channel or a competitor’s reaction. Once you name it, you know where the review must begin.

The second step is contact with reality. An assumption about prices can be checked against a real price list, an assumption about customer needs against a conversation, an assumption about margins against a calculation of your own. AI can help prepare these checks, gather counter-arguments and make gaps visible. But the outcome of the review must be owned by someone who knows the company.

Fit rather than patterns

AI systems are good at transferring known business model patterns to a new situation. That is useful because it opens up the search space. But it is also the point at which drafts most easily become generic. A pattern that works for some company does not yet answer whether it fits the capabilities, customer relationships and resources of this company.

It is therefore worthwhile for teams to confront each proposed building block with a simple counter-question: what does this model require that we can already do, and what would we first have to build? Where this question finds no clear answer, that is usually where the real strategic work lies.

Reviewing as shared work

When a draft is produced in minutes, the phase in which understanding used to build up is missing. This phase has to be reintroduced deliberately – not as generating, but as reviewing. That can mean a team taking an AI draft apart together, discussing the assumptions one by one and recording how each will be tested.

This includes the willingness to be wrong. Anyone reviewing a draft must be open to the possibility that it does not hold up. A team that cultivates this openness genuinely gains speed with AI. A team that does not mainly gains more drafts.

Key takeaways for leadership teams

• A high AI share of the output says nothing about the quality of a business model.
• The decisive part of the work is proving that a draft holds up.
• Every model should be examined for the one assumption that carries it.
• This review requires contact with reality and someone who knows the company.

If you want to know whether an existing strategy holds up, the Strategy Check makes assumptions visible and alternatives comparable. Go to the Strategy Check →

Source: Original on erwinhettich.com