Decisions, not workflows

Most leadership teams cannot, off the top of their heads, list the decisions their company makes that matter. They can list functions. They can list initiatives. They can list KPIs. The decisions, the recurring choices that determine what the company produces, are usually invisible to the people running the company, because they have been made by the same people for so long that they have stopped looking like decisions and started looking like the way things are.

This is the first thing I do in any engagement. We sit down for an afternoon and surface the decisions. The conversation is uncomfortable in roughly the same way every time, which makes me think the discomfort is structural.

Here is how it goes.

I ask: what are the five most consequential recurring choices your company makes? Not events, not projects. Choices that recur, that shape what you sell or how you sell it.

The first three answers come fast. Pricing. Hiring. Roadmap prioritization. Standard.

The next two are harder. People reach. They name things that are not quite recurring, or not quite consequential, or not quite a choice. Someone says "channel mix." Someone says "which leads to call back first." Someone says "which projects to greenlight." We write them all down.

Then I ask the second question, which is the one that does the work.

For each decision, who actually makes it? Not who is supposed to. Who actually does.

This is where the room gets quiet.

It turns out that the most consequential choices are usually being made by accident, by accretion, or by whoever is closest to the keyboard at the moment of the choice. Pricing in a B2B company is "made" by sales reps in the moment, inside an envelope nobody has revisited in two years. Hiring is "made" by the loudest hiring manager. Roadmap prioritization is "made" by whoever spoke last in the most recent product meeting.

I want to be careful here. This is not a critique of the people involved. It is a structural fact about most companies. Decisions of this type are usually distributed, partially formalized, and made under time pressure with whatever signal happens to be at hand. They become invisible because they are made constantly, by many people, against ambient assumptions nobody wrote down.

The reason this matters for AI strategy is that these distributed, half-formal, ambient-signal decisions are precisely the ones AI is well suited to change. Not by automating them away, in most cases. By making them legible. By bringing them into a single surface where they can be seen, calibrated, and made consistently.

Once a leadership team has its list of five decisions, the next step is to ask, for each one, three questions:

The first is whether this decision is currently underweighted or overweighted in management attention. Most companies overweight one decision and underweight the other four. The CEO has spent two years on pricing. Hiring has gotten almost no design.

The second is what signal currently feeds the decision. The honest answer is usually "less than we would like." Pricing in most B2B companies feeds on whatever the rep happens to know about the prospect. Hiring feeds on resumes and gut feel. Roadmap prioritization feeds on whoever spoke last. Each of these has a much richer signal source available now, in 2026, that did not exist in 2022.

The third is what would change in the company if the decision were tightened by 50 percent. This is the question that produces the strategic insight, because it forces the team to imagine the second-order consequences. If pricing were 50 percent more consistent, what happens to the gross margin distribution across customers? What happens to the win rate on the bottom quartile of deals? What happens to the sales cycle on the top quartile? The answers are usually very interesting and almost never on the current dashboard.

What you have at the end of an afternoon, if it goes well, is a map. Five decisions. For each, who really makes it, what signal feeds it, and what would change if it were tightened. From the map, the AI strategy almost writes itself: pick the one or two decisions where the leverage is highest and the signal is most available. Build the surface for those decisions first. Let the workflows reorganize around them.

This is unfashionable advice. The fashionable advice is to map your processes, score them on impact and effort, and run pilots on the top quartile. I do not think that advice is wrong, exactly. It is just answering a smaller question. It tells you how to make existing activities cheaper. It does not tell you what business you want to be in.

The decision map tells you what business you want to be in. It is, in my experience, the single most useful artifact a leadership team can produce in the first month of taking AI seriously.

A small note on what this is not. It is not a transformation framework. It is not a maturity model. It is not a six-quarter roadmap. It is one afternoon, five rows in a table, and a difficult set of follow-up conversations. The afternoon is the cheap part. The conversations are the expensive part. Most teams skip the conversations, which is why most teams end up with workflow projects.

I have come to think the test of whether a company is taking AI seriously is not the size of the budget or the number of pilots. It is whether they have done this exercise honestly. Most have not. The ones that have are easy to spot, because they talk about their AI work in a different vocabulary than everyone else. They do not say "we are deploying AI in support." They say "we are tightening our routing decision." It is a small linguistic difference. It indicates an enormous strategic one.