Choosing What to Fund, AI Investment as a Portfolio Decision

One of the easiest ways to make a weak investment decision is to evaluate every opportunity one at a time. A compelling use case appears. A strong sponsor supports it. The efficiency case looks attractive. A competitor is doing something similar. Viewed in isolation, each opportunity can seem reasonable. The difficulty appears when all of them compete for the same money, people, data, technology, and leadership attention.

I encountered this while independently developing a model for allocating AI investment across a multi-product financial-services environment. The question was not whether any single use case had value. Several did.

The harder question was: How should an enterprise decide which AI opportunities deserve investment when several appear valuable at the same time?

That changed the problem. It was no longer enough to rank ideas individually.

The real question became: Which combination of investments gives the organization the best chance of creating real value?

Good Ideas Still Compete for Finite Capacity

Organizations rarely face a shortage of ideas. They face a shortage of capacity. An AI opportunity may have a strong customer case. Another may reduce operating cost. Another may address a regulatory need. Another may create a reusable capability that several products could depend on later. Each can be strategically reasonable. But they cannot all move first.

This is where prioritization becomes more difficult than scoring. A list of individually attractive initiatives does not tell leaders what the portfolio should look like. Investment choices have to account for competition between opportunities as well as the merits of each one. That means considering more than potential value:

• What would funding this prevent us from funding?
• What has to exist before it can succeed?
• What does it enable later?
• What happens if we delay it?

Those are portfolio questions.

Value and Readiness Are Different Questions

One of the most useful distinctions in the model was separating value from readiness. They often get blended together. A high-value initiative can look like an obvious investment. But an initiative can be strategically attractive and still be difficult to pursue now. The data may not be ready. A required platform may not exist. The workflow may still be unclear. The organization may lack the operational capacity to support the change. Another opportunity may be easier to launch but offer relatively little lasting value. That creates an important tension:

High value does not always mean invest now. High readiness does not always mean invest at all.

Those are different decisions. Value helps answer whether an opportunity is worth pursuing. Readiness helps answer when and how it can realistically move. Treating them separately makes the tradeoff more visible. It also prevents organizations from confusing ease with importance.

Dependencies Can Matter More Than Scores

A scoring model can create useful discipline. It can make assumptions visible and force teams to compare initiatives using common criteria rather than advocacy alone. But a score is only part of the decision.

The score starts the conversation. Dependencies change it.

Suppose Initiative A scores highest and Initiative B scores slightly lower. On the surface, A should go first. But what if A depends on a capability that B creates? What if B establishes shared data, integration, identity, governance, or workflow infrastructure that several later initiatives can reuse? What if funding B first makes A cheaper, faster, or safer later? At that point, the portfolio decision changes.

The value of an initiative is partly contained in its own business case. It can also come from what it enables elsewhere. This is why foundational capabilities can be easy to undervalue. They may not have the most exciting standalone story, and their return may be distributed across several products or future initiatives. But without them, the organization may repeatedly solve the same problem, build duplicate capabilities, or delay higher-value work. A portfolio view therefore has to consider more than direct value.

It has to consider enablement value. Useful questions include:

• Which investments make other investments possible?
• Which capabilities can be reused?
• Which dependencies create sequencing constraints?
• Which choices reduce duplicated effort later?

Those answers can materially change what deserves to move first.

Risk Should Change the Investment Approach

Risk belongs in the investment decision. But it should not automatically function as a stop sign. A higher-risk initiative may still be worth pursuing if the potential value is strong. The investment approach simply needs to reflect the uncertainty.

That may mean:

• funding a smaller experiment first;
• narrowing the initial scope;
• requiring stronger evidence before the next commitment;
• building a dependency before funding the full initiative;
• adding specific controls before scale;
• delaying the investment until conditions improve.

This is where risk becomes an investment variable rather than a separate governance discussion.

The practical question is: How should this risk affect the amount, timing, sequence, or conditions of the investment?

That keeps risk connected to the funding decision.

Prioritizing, Sequencing, and Funding Are Different Decisions

These terms are often used as though they mean the same thing. They do not.

• Prioritizing asks which initiatives matter most.
• Sequencing asks what should happen first, second, or later.
• Funding asks what level of commitment should be made now.

An initiative can rank highly and still not be first. Another can deserve early funding because it removes a dependency even if its standalone value is lower. A third may deserve limited funding now while the organization learns enough to decide whether a larger commitment is justified. Keeping those decisions separate creates more flexibility. It also helps avoid a common trap: treating the prioritization list as if it were already the investment plan.

A ranked list is not a portfolio.

The portfolio emerges when leaders decide how priorities, dependencies, timing, and funding fit together.

Investment Discipline Requires Visible Assumptions

Portfolio decisions are rarely purely analytical. Different leaders will value different things. One may emphasize near-term efficiency. Another may prioritize customer impact. Another may care most about regulatory exposure, strategic differentiation, or platform leverage. That is normal. A structured investment process should make those assumptions visible enough to discuss.

Questions worth making explicit include:

• Why does this initiative rank higher?
• What evidence supports the value case?
• What has to be true for the expected benefit to materialize?
• What dependencies are we assuming will be solved?
• What would cause us to reduce, delay, or stop the investment?

Those questions make the decision easier to revisit later. They also reduce the risk that funding becomes a contest between the strongest advocates. The goal is not to replace leadership judgment with a scoring model.

The goal is not to remove judgment. It is to give judgment better structure.

The Best Portfolio Is Not the Collection of the Best Individual Ideas

This is the part of portfolio thinking I find most useful. If every initiative is evaluated independently, the logical outcome is to fund the highest-scoring ideas. But the strongest portfolio may look different. It may include a foundational investment with modest direct value because several future initiatives depend on it. It may delay an exciting opportunity because the organization is not ready to support it. It may fund a smaller experiment rather than a full program because uncertainty is still high. It may also stop an initiative that once looked attractive because another investment now creates more value.

The portfolio is a system of choices. Those choices interact.

That means the real questions are broader than: Which idea has the best score?

They are closer to:

• What creates meaningful value?
• What is actually ready?
• What dependencies or shared capabilities matter?
• What level and timing of investment make sense now?
• What do we need to learn before committing more?

No scoring system can answer those questions automatically. Nor should it.

Investment decisions require judgment because organizations are choosing among different types of value, different timelines, and different levels of uncertainty. A good portfolio process makes that judgment more informed, more visible, and easier to revisit. The goal is not to identify the highest-scoring AI idea.

The goal is to make a coherent set of investments that the organization can actually support, sequence, and turn into value.