Decision Memo: Do Not Fund AI. Fund the Problem Exposure.
AI belongs in the innovation portfolio. It should not become the portfolio logic.
Tuesday’s article made the case that the innovation economy is solving the wrong problems. That 61 percent of global venture capital flowing into AI is not a portfolio — it is a concentration. And that we are scaling computational power faster than the human systems required to absorb it.
This is the decision memo for that argument.
AI is the easiest budget to approve right now.
It signals ambition. It reassures boards that the company is awake. It gives executives a funding category that feels current, defensible, and nearly impossible to challenge without sounding like you just got back from a long sabbatical.
That does not make AI investment wrong.
AI deserves serious attention. It behaves like electricity or the internet once did: pervasive, still improving, and capable of enabling change not just in one product or process but across entire operating models. The technology matters less as a tool and more as a platform or even as general infrastructure. Everything built on top of it is where the value actually moves.
That is exactly why AI is strategically important. It is also why AI is dangerous as a portfolio category.
Because AI can be applied almost everywhere, it can justify almost anything. The board asks for an AI strategy. The executive team asks for AI use cases. Functions scan for automation potential. Innovation teams quietly repackage existing projects around generative AI, agents, productivity, data, or internal efficiency.
Soon the portfolio contains AI activity.
Activity is not allocation.
The stronger question is not whether the company invests enough in AI. It is which problem exposures deserve capital, and where AI is the best available lever to reduce them.
This distinction matters. Productivity gains do not appear simply because a technology is powerful. They appear after complementary innovations, process redesign, and real organizational absorption. That gap between adoption and absorption is where most companies currently sit.
The risk is not that AI is irrelevant. The risk is mistaking technical adoption for organizational change.
There is already evidence that AI creates measurable gains in specific contexts. One large field study with customer support agents found that access to a generative AI assistant increased productivity by 14 percent on average. But the same study showed strong variation across individuals.
That variation is the point.
AI does not improve “work” in general. It improves specific tasks under specific conditions. Researchers describe this as a jagged technological frontier: AI helps strongly in some areas, and can actively reduce quality in others when applied beyond its effective boundary.
For leaders, the implication is straightforward. AI funding should not start with the tool. It should start with exposure.
This is where the Horizon 1, Horizon 2, Horizon 3 model fails — and why it was always a bet disguised as a framework.
The horizon model places technologies into time buckets. H1 is what runs the business today. H2 is what builds the business next. H3 is what could reinvent it eventually. Clean, intuitive, easy to present on a slide.
The problem is that it rests on an assumption no one can actually justify: that you know how fast a technology will diffuse and how hard it will hit your business. Placing AI in H2 or H3 a few years ago was not analysis. It was a perspective. A guess with a label on it.
The latest AI wave exposed this directly. What most companies assigned to H3 — ambient AI, generative interfaces, cognitive automation — arrived in H1 before the portfolio review was finished. The model did not fail because the horizon logic is wrong in theory. It failed because diffusion speed is not yours to decide. Markets, competitors, customers, and cost curves decide it. Your horizon assumption is just how long you think you have. And that assumption is now mostly wrong.
A technology-centered portfolio forces you to bet on timing you cannot control.
A better lens does not ask where a technology sits on a diffusion curve. It asks where the company is exposed — now, next, and later — regardless of which technology turns out to be the relevant lever.
Short-term exposure is visible pressure on current performance: cost, quality, speed, capacity, compliance, service load, or operational bottlenecks. Whatever is already hurting.
Mid-term exposure is a capability gap that is forming but not yet fully painful. Changing customer expectations. Competitors quietly improving their unit economics. Internal tool adoption that signals a skills gap already opening.
Long-term exposure is structural relevance. Market boundaries blur. Outsiders become credible. Value migrates from products to outcomes. Business models are reshaped by cheaper cognitive work.
A serious portfolio holds all three at once. Not equally. Deliberately.
The goal is not to spread resources across time horizons because the model looks balanced. The goal is to allocate where learning would change a material decision — and to stop confusing a view on technology timing with a view on actual business risk.
The AI Innovation Portfolio Checklist
Use this before approving, extending, or scaling AI-related innovation funding.
