The three diamonds of AI value.

Most product design people know the Double Diamond concept. First you work in the problem space: you diverge to discover the problems, then converge to define the right problem. Then with the chosen problem you go to the solution space: you develop many possible solutions for the problem, and then again you choose the best one to deliver the right solution. And then you iterate.

In 2026 that map is missing a part that now matters most.

Buildable was never the same as valuable

Companies are no longer asking whether to do AI. Not even where or how to do it. They've most probably done a lot of learning, run the studies, bought the tools, tried the pilots; or they do it now, as we speak. FOMO is a powerful driver. And yet the results aren't showing up where it counts. MIT's NANDA research found that roughly 95% of enterprise GenAI pilots deliver no measurable P&L impact. Gartner expects about a third of GenAI projects started in 2024 to be abandoned by the end of 2026. The spend is real, problems get identified and solutions are delivered. The value mostly isn't.

The reason is simple, and it's not technical: a solution being buildable (and even working nicely) says nothing about whether it creates value in the real conditions of your business. That gap is invisible on a two-diamond map. So let's add a third one.

Three diamonds

1 ◆ Problem space: which problems are worth AI? Diverge across the problems and outcomes that actually matter to the business; converge on the few that are both high-value and genuinely AI-suited. Business first, AI second, not the reverse. You probably know already many problems itching for years, where you have almost given up, realising that solving it "is impossible" it may require very complex new technology or maybe scaling up expensive manual labour - both can now be challenged with smart AI application.

2 ◆ AI solution space: which solution actually works? Diverge across technologies, approaches, data, and tools; converge on one technically validated pilot, testable increment for business use.

3 ◆ Value-realization space: now you test it in scale: does it move the P&L? Diverge across the ways a working solution could be deployed, adopted, and embedded into the operating model; converge on the configuration that actually shifts the numbers in the desired direction, or the decision to stop.

The third diamond is where 2026 is being won and lost. Treating it as an afterthought, or being too slow here is exactly why so much AI spend evaporates. The key infrastructure here is a real-time business data: collection, processing and analyzing capability: good if you have cross-cutting up-to-date weekly business metrics, connected to the AI initiatives. Even better if you have it in daily or even more real-time pace, depending on the business specifics. Forget the manual Excels and PowerPoints that business analysts polish - that may have worked before when business was finetuned quarterly or even yearly. Now you need proper backend data stack: descriptive, diagnostic, predictive and prescriptive analytics. The prescription is: does that AI initiative work or it just needs to be axed.

Mind the valleys

Organizations rarely fail inside a diamond. They fall into the gaps between them.

The gap between problem and solution is supposed to be one and clean problem to be solved. In reality we often see here Opportunity theatre - endless opportunity maps and workshops, nothing actually committed. There are indeed always too many problems on the table, unsure which one is really most important and where an AI initiative could help. Here again being diligent (and capable) on data front helps a lot: organizations should really see from data, and not guesswork and guts feeling where are really the most immediate problems.

Between solution and value there should be one specific focused delivery which works. Instead we see Pilot purgatory - a pilot (or even many) that works impressively in the demo, but never industrializes into the business.

If your AI program is stuck, it's almost certainly in one of these valleys, not in the technology.

Where are you on the map?

Before spending another euro, locate yourself:

• Not started: you're entering Diamond 1.
• Opportunities mapped, nothing built: you're stuck at the first valley.
• Built and piloting: you're in Diamond 2, with the second valley ahead.
• Spending, but no proven return: you're in Diamond 3, where the honest questions live.

Knowing the stage tells you what to do next, and what help is even worth buying.

A note from the author

An outside advisor cannot do equal cut of every diamond here, and you should be suspicious of anyone who claims to. I believe that in the problem space, an independent partner adds the most: cross-company pattern recognition and the freedom to say "not this." In the solution space, the outside value is in validation and the go / no-go, and the actual build belongs to delivery teams which sit closest to the real operations in-house. In the value-realization space, the honest contribution is a verdict (is value realizable here, what should you scale, and what should you stop); no one can promise that someone external can make adoption happen for you.

The goal was never more software, or more AI. It's better decisions and realized value; or a clean, early stop. Build less. Learn faster. And know which diamond you're standing in.