AI in three Circles

Claas
2 days ago
3 min read

How most AI initiatives end up incomplete

You know Don McMillan? An engineer who became a comedian and explains things with venn diagrams. The result is a very funny way of explaining different combinations. Slightly exaggerated, but close enough to reality that you recognize it immediately..

I like that format, so I tried something similar for AI initiatives.

In most organisations, three elements keep showing up in different forms: an AI strategy, data (imrovement) initiatives and use cases in production. In theory, they belong together. In practice, they rarely appear at the same time.

I

f you map this in a simple diagram, the combinations are quite familiar.

If you have strategy and data, you are… running workshops.
If you have strategy and production, you are… showing a demo.
If you have data and production, you are… solving something locally.
If you have all three, you are… the unicorn.

Most organisations will recognise themselves in one of the first three.

Strategy and Data

This is often where things start. There is a clear ambition, data is being collected and structured and there is no shortage of ideas about what could be done with it. From the outside, this looks like progress and in many ways it also is.

At the same time, very little reaches production. Use cases remain conceptual, value is discussed but not measured, and ownership is spread across teams in a way that makes coordination easier but decisions harder. The organisation stays active, producing roadmaps, running workshops and aligning stakeholders, while the step into something concrete is repeatedly postponed.

What tends to hold this back is not capability, but commitment. As long as things remain abstract, agreement is easy. The moment a specific use case is chosen, trade-offs become visible and priorities need to be set.

Moving forward here usually requires less expansion and more focus. Narrowing down to one use case, defining what success looks like and assigning clear ownership often helps to change the dynamic.

Strategy and Production

In this situation, something is already running. There is a use case in production that can be demonstrated, which creates visibility and often reinforces the overall narrative that progress is being made.

Looking closer, these solutions are frequently supported by a significant amount of invisible effort. Data is prepared manually, pipelines are adjusted on the fly and a small number of people ensure that everything keeps working. From the outside, it appears stable, but it depends on conditions that are difficult to reproduce.

This pattern is not unusual. Building something that works once is relatively straightforward compared to building something that works reliably and at scale. The underlying challenge sits in the foundation the solution depends on, particularly the availability and quality of data.

Progress in this situation usually starts when these dependencies become visible. As long as temporary fixes remain hidden, there is little incentive to address them. Once they are exposed, the focus shifts toward making the system more robust.

Data and Production

This combination is often the closest to something that creates real value. There is data available, and there are solutions that use it in a meaningful way. In many cases, these solutions emerge from teams close to the problem, built pragmatically without waiting for a broader narrative.

Because of that, they tend to be effective, but also limited in reach. They remain local, are not widely visible and are not actively connected to a larger context. What works in one place does not automatically spread to others.

When these solutions are eventually discovered, they are often turned into broader initiatives. That can help with scaling, but it also changes how they are managed and can slow down what originally made them effective.

The challenge here is not to rebuild what already works, but to make it visible and reusable without overcomplicating it.

The full picture

Most organisations have all three elements somewhere. Strategy exists, data exists, and there are solutions in production. They are just not aligned in a way that allows them to reinforce each other.

Bringing them together is about connecting what is already there. No new elements required. It equires decisions, clear ownership, and timing across different parts of the organisation.

When that happens, things tend to look simpler than expected, not because the problem is simple, but because the pieces finally fit.

That is what makes the unicorn rare.

AI in three Circles

How most AI initiatives end up incomplete

You know Don McMillan? An engineer who became a comedian and explains things with venn diagrams. The result is a very funny way of explaining different combinations. Slightly exaggerated, but close enough to reality that you recognize it immediately..

I like that format, so I tried something similar for AI initiatives.

In most organisations, three elements keep showing up in different forms: an AI strategy, data (imrovement) initiatives and use cases in production. In theory, they belong together. In practice, they rarely appear at the same time.

I

f you map this in a simple diagram, the combinations are quite familiar.

If you have strategy and data, you are… running workshops.

If you have strategy and production, you are… showing a demo.

If you have data and production, you are… solving something locally.

If you have all three, you are… the unicorn.

Most organisations will recognise themselves in one of the first three.

Strategy and Data

This is often where things start. There is a clear ambition, data is being collected and structured and there is no shortage of ideas about what could be done with it. From the outside, this looks like progress and in many ways it also is.

What tends to hold this back is not capability, but commitment. As long as things remain abstract, agreement is easy. The moment a specific use case is chosen, trade-offs become visible and priorities need to be set.

Moving forward here usually requires less expansion and more focus. Narrowing down to one use case, defining what success looks like and assigning clear ownership often helps to change the dynamic.

Strategy and Production

In this situation, something is already running. There is a use case in production that can be demonstrated, which creates visibility and often reinforces the overall narrative that progress is being made.

This pattern is not unusual. Building something that works once is relatively straightforward compared to building something that works reliably and at scale. The underlying challenge sits in the foundation the solution depends on, particularly the availability and quality of data.

Progress in this situation usually starts when these dependencies become visible. As long as temporary fixes remain hidden, there is little incentive to address them. Once they are exposed, the focus shifts toward making the system more robust.

Data and Production

This combination is often the closest to something that creates real value. There is data available, and there are solutions that use it in a meaningful way. In many cases, these solutions emerge from teams close to the problem, built pragmatically without waiting for a broader narrative.

Because of that, they tend to be effective, but also limited in reach. They remain local, are not widely visible and are not actively connected to a larger context. What works in one place does not automatically spread to others.

When these solutions are eventually discovered, they are often turned into broader initiatives. That can help with scaling, but it also changes how they are managed and can slow down what originally made them effective.

The challenge here is not to rebuild what already works, but to make it visible and reusable without overcomplicating it.

The full picture

Most organisations have all three elements somewhere. Strategy exists, data exists, and there are solutions in production. They are just not aligned in a way that allows them to reinforce each other.

Bringing them together is about connecting what is already there. No new elements required. It equires decisions, clear ownership, and timing across different parts of the organisation.

When that happens, things tend to look simpler than expected, not because the problem is simple, but because the pieces finally fit.

That is what makes the unicorn rare.

Recent Posts

Comments