The opening was not on the use-case list

Meta description: AI opportunity lists can miss the business constraint when they catalogue tasks rather than the relationships shaping performance.

Miklos Roth

I have read a lot of AI opportunity maps. They are thorough documents — workshops held, ideas harvested, use cases scored and heatmapped, quick wins circled in a color that suggests confidence. And in the versions of this story that stay with me, the map missed the thing that mattered. The opening sat in a back-office joint nobody sponsored, in a constraint so unglamorous that nobody brought it to the workshop. The company was not short of ideas. It was blind in a specific, repeatable place.

Imagine a management team compiling forty AI use cases. The list is full of sensible ideas: draft customer communications, generate sales material, summarize reports, speed up support. Every department has contributed. The list may be useful, but the list itself does not establish where the greatest change is possible. A problem that sits between departments may be absent because no single manager owns it.

What do opportunity maps systematically include?

The maps I read are consistent in what they capture: visible pain, visible customers, visible sponsors. Ideas surface when someone with a seat at the workshop feels the pain and can imagine the tool. This is a reasonable way to run a workshop, and it builds the blind spot directly into the output. The map shows where the company looked, not where the value is. I have learned to read an opportunity map as a document about the company’s attention, and to ask what the attention was aimed away from.

Why do visible use cases crowd out the constraint?

My Roth Complexity research attempts to bring observations from different disciplines into a shared analytical language. In that work I use the S-I-C-T framework to examine interacting dimensions of complex systems. For a company, the interest is not in placing an abstract label on its problems. The interest is whether relevant dimensions can be made observable with suitable proxies.

The critical word is suitable. A measure that makes sense in one business may be irrelevant in another. A useful proxy has to correspond to a feature of the actual organization and be tested against meaningful outcomes. Examining these relationships may reveal gaps in the system’s capacity to change. These are hypotheses to investigate, not opportunities that a formula has already proved.

What I can describe is the crowding itself. A map with forty use cases feels complete; it absorbs the planning energy. Each use case has a sponsor who defends it, a score that justifies it, a place on a roadmap. The constraint — the place where the business is actually limited — tends to have none of these. No sponsor, because it sits between functions. No score, because scoring models reward what is easy to describe. The map did not overlook the constraint. The mapping process was not built to register it.

What kind of opening has no internal owner?

Read more: Miklos Roth.

The false explanation in these companies is that they need more ideation — another workshop, a suggestion channel, an innovation challenge. I have watched ideation volume double while the same constraint sat untouched. The opening that matters, in the pattern I keep seeing, is ownerless by construction: it lives at a boundary, it benefits everyone and belongs to no one’s P&L, and pursuing it would require two departments to share credit. Ownerless openings do not lack value. They lack a constituency.

What pattern marks a hidden opening versus a pet project?

Pet projects are easy to spot in hindsight: they have an enthusiastic sponsor, a demo that impresses, and a value story that requires faith. Hidden openings look the opposite — unglamorous, cross-functional, attached to a metric the leadership team claims to care about and never reviews. I have watched companies fund the pet project and starve the opening in the same budget cycle, and the scoring model rated the pet project higher, because pet projects are built to be scored and openings are not.

A fashionable project often begins with a capability: a model can write, classify, search, or summarize, so a team looks for a place to use it. A potentially valuable opportunity begins with a meaningful business problem and asks what prevents improvement. That distinction changes which evidence matters.

If every team improves its own output but downstream work becomes harder, a portfolio may create activity without cohesion. A serious review must see how local gains interact with the rest of the organization and with the outcome the organization actually wants.

I approach this as a question about the organization as a complex system. A company’s performance is not simply the sum of improvements made to individual tasks. It also depends on how its parts interact: what information moves, which relationships support the work, and where a potential transformation becomes blocked.

What question makes the blind spot discussable?

The difficulty is not analytical. In most of these companies, someone knows where the constraint is — usually several someones, in operations, who have known for years. What is missing is a way for that knowledge to enter a leadership conversation without sounding like a complaint. There is a question that opens that door, and I have noticed it is almost never on the workshop agenda.

The open question is not how many AI uses a company can imagine. It is which important transformation remains invisible because the organization is measuring tasks instead of the system connecting them?

PR excerpt (80–120 words, Miklos’s own words): Imagine a management team compiling forty AI use cases. Every department has contributed, yet the list itself does not establish where the greatest change is possible — a problem that sits between departments may be absent because no single manager owns it. A fashionable project begins with a capability and looks for a place to use it; a valuable opportunity begins with a meaningful business problem and asks what prevents improvement. The open question is which important transformation remains invisible because the organization is measuring tasks instead of the system connecting them.