Decision Foundations in the Age of AI

What a decision-making foundation actually is

A decision-making foundation is a clear overview of requirements, options and risks that forms the basis for a well-founded decision, often supported by a value-benefit analysis and a clearly structured presentation for management, C-level or the board of directors.

At first, this definition sounds like a calculation process. Gathering requirements, evaluating options, weighting them, presenting a result. This part can genuinely be sped up with AI support. What the definition assumes, however, is something else: that the requirements and options are already known. That is exactly where the real work begins, and exactly where the difference lies between an evaluation and a decision-making foundation.

The actual gap

A model can work with the data it is given. It cannot know which concerns are never voiced in the boardroom, yet still influence the outcome. It does not know the history between two departments that makes an otherwise sound option politically impossible. It cannot judge who needs to be brought on board first, so that a decision is actually supported afterwards, not just passed.

And it carries no responsibility. A decision-making foundation that holds up in front of a management team or a board of directors does not stand on its own. Someone has developed it, represents it, can answer questions in the room, and stands behind it with their own name. That is not a minor detail, it is the core of what outside support delivers that a generated document cannot.

Where we use AI ourselves

This does not mean AI has no place in this work. We use it ourselves, deliberately, wherever it makes the work faster, better founded or more precise, in preparing data, structuring initial options, checking assumptions. This gains speed, and often thoroughness too, because more options can be worked through in less time. Judging which of these options actually makes sense in the specific case, and the recommendation that results from it, stays with us. So does the responsibility for it.

An example from practice: site evaluation

What this looks like in practice is well illustrated by a site evaluation we supported a company through. At the start, a concrete economic goal was known. What was not known: the requirements for a new site, the possibilities actually available, the realistic options. All of this first had to be uncovered and defined before anything could even be compared.

A model would have had nothing to work with at this stage, because there were no options yet for it to evaluate. The options only emerged, iteratively, in meetings with the CEO, the group CEO and the owner. Each of these meetings narrowed the possibilities further, depending on what was practically feasible and which spaces were actually available. The final solution stood at the end of a process, not at the start of a calculation.

That is the part no software can take over. Not the calculating itself, but working out what needs to be calculated in the first place, in conversation with the people who will ultimately live with the solution and be responsible for it.

What this means for your next decision

The bigger a decision, the more this distinction is worth keeping in mind. AI tools help prepare a decision-making foundation faster and on broader ground. They do not replace the conversation in which the actual options first appear, and they take on no responsibility for the recommendation that ends up on the table.