Executive Summary
AI adoption is widespread. Enterprise-level business impact is not. The gap between the two has many causes, but one of the most consistent and least discussed is this: most organizations are managing AI as a collection of one-off projects rather than as a deliberately constructed portfolio, with no principled way to decide which initiatives to pursue, in what order, and why. Key insights include:
- One-off project thinking actively destroys value through resource cannibalization, blocking effects, and missed synergies that are invisible at the project level
- The right unit of analysis for AI is not the individual project. It is the portfolio, a deliberately selected set of initiatives that together align to the organization’s strategic priorities and maximize the value AI can deliver
- Moving from one-off thinking to portfolio thinking requires an AI project prioritization methodology, a structured and repeatable process for surfacing the right projects and assembling a mix that is coherent and defensible
- Organizations that make this shift stop accumulating AI projects and start building AI capability, and that distinction is what enterprise impact actually requires
In a previous post, we noted that AI is everywhere. Enterprise impact isn’t. The research is unambiguous. According to McKinsey’s 2025 State of AI report, 88 percent of organizations now use AI in at least one business function. Yet only 31 percent have scaled it enterprise-wide, and only 39 percent report any measurable effect on enterprise-level performance. PwC’s 2026 Global CEO Survey found that 56 percent of chief executives saw neither higher revenues nor lower costs from their AI investments.
The instinct is to look at the technology. A better model, a more sophisticated use case, a different vendor. But in most organizations, the technology is not the problem. Something structural is missing. And one of the most consistent structural gaps we see is this: organizations are managing AI as a collection of one-off projects. Not as a portfolio. Not with a principled methodology for deciding which initiatives to pursue, in what order, and why.
Until that changes, the pattern repeats. More pilots. More activity. Limited impact.
The Hidden Cost of One-Off Project Thinking
Ask most AI leaders how their organization selects AI projects and the answer, if they are honest, sounds something like this. Department heads surface ideas. Some get executive sponsorship. Some generate pilot results that create momentum. Others get quietly shelved when a louder priority arrives. There is no consistent framework. No shared criteria. No visibility across the whole.
This feels like an AI project prioritization problem. It is actually something more serious. One-off project thinking does not just produce poor prioritization. It actively destroys value in three distinct ways that are almost impossible to see until you step back and look at the initiative landscape as a whole.
The first is resource cannibalization. Organizations operate with a constrained pool of budget, AI talent, data engineering capacity, and IT integration bandwidth. When projects compete for those resources individually, each making its own case, to its own sponsor, on its own timeline, the project that wins is often not the highest-impact one. It is the one with the most persuasive advocate or the most political momentum. Resources flow toward whoever argues loudest, not wherever they would create the most value.
The second is blocking effects. A project that gets greenlit early, perhaps because it had vocal sponsorship or a quick pilot result, can consume the organization’s limited integration capacity or data engineering bandwidth ahead of initiatives with far greater strategic value. By the time the sequencing problem becomes visible, months and budget are gone. The blocked projects have lost their window. There is no vantage point from which to see it coming, because each project is being evaluated on its own terms with no visibility into what it might be crowding out.
The third failure mode is the most consequential, and the least obvious: missed synergies. When projects are evaluated in isolation, each one must justify itself on its own merits. But some of the most valuable AI initiatives do not look compelling when viewed alone. Consider a shared data pipeline that no single business unit can justify funding independently. Viewed in isolation, it looks like overhead with no clear ROI. Viewed across the portfolio, it is the foundation that makes three other projects feasible that were previously blocked by data access or quality issues, projects that together might deliver significant business value but could not have been greenlit without it. The synergy is real. But it is only visible from the portfolio level. One-off project thinking is structurally blind to it.
The same logic applies to shared model infrastructure, common integration layers, and reusable AI services. Projects that each appear marginal or technically infeasible on their own may become highly executable and highly valuable when they share foundational investments. The opportunity compounds. But only if someone is looking at the whole.
These are not edge cases. They are the normal operating conditions of any organization running multiple AI initiatives without a coordinating framework. And they are a significant part of the reason why AI activity does not compound into enterprise impact.
The AI Portfolio as the Strategic Answer
The solution is not to evaluate projects more carefully in isolation. It is to change the unit of analysis entirely, from the individual project to the portfolio.
Think of it like a mutual fund. A fund manager does not simply pick the best individual stocks and call it a strategy. They construct a deliberate collection of holdings across asset classes, risk profiles, and time horizons that together maximize return while managing risk. No single holding tells the whole story. The value is in the composition. A portfolio that is all high-risk growth bets is as poorly constructed as one that is all conservative income holdings. The skill is in the mix.
An AI portfolio works the same way. It is a deliberately selected set of initiatives that together align to the organization’s strategic priorities, maximize the value AI can deliver to the business, and manage risk across the portfolio as a whole. Some initiatives deliver near-term returns that build credibility and fund longer-horizon bets. Others build capabilities that compound over time. Others create the shared infrastructure that makes everything else possible. No single initiative does all of this. The portfolio does.
This is how organizations close the gap between AI activity and enterprise impact. Not by finding better individual projects. By assembling the right mix.
At Strategy of Things, we organize AI initiatives across five portfolio archetypes, the asset classes of a well-constructed AI project portfolio.
Big Bets are high-risk, high-reward transformational plays. These are initiatives with the potential to reshape a business model, create an entirely new source of competitive advantage, or fundamentally change how the organization operates. They require patience, tolerance for uncertainty, and strong executive sponsorship. The outcome is not guaranteed, but the potential upside justifies the risk when balanced appropriately against the rest of the portfolio.
Strategic investments are those that directly align to the organization’s most important strategic priorities. These are not optional bets on future upside. They are initiatives the organization cannot afford to deprioritize. The cost of inaction is real: falling behind competitors, losing critical capability, or undermining the organization’s ability to remain viable in its market. Strategic investments may take time to deliver their full value, but their connection to what matters most to the business makes them non-negotiable.
Quick Wins are fast, visible initiatives that deliver meaningful returns in a short timeframe. Their value is not just in the outcome. It is in the momentum they create. Quick Wins build organizational confidence in AI, demonstrate to leadership and the board that investment is translating into results, and maintain credibility while longer-horizon bets develop. Every portfolio needs them, not because they are the most strategically significant initiatives, but because they keep the organization moving and believing.
Low Risk projects are highly doable initiatives where the conditions for success are already in place. The risk of failure is low because the prerequisites are met: the technology is mature and well understood, the data is clean and accessible, and the implementation does not demand significant organizational change or technical complexity. They deliver steady, incremental value and keep the portfolio productive without demanding disproportionate resources or leadership attention.
Foundational investments often appear unattractive when evaluated in isolation. The investment can be significant, and the direct return modest compared to other AI initiatives competing for the same budget. But this single-project view misses the multiplier effect entirely. What Foundational projects deliver is the shared infrastructure upon which multiple other projects depend: unified data pipelines, common model hosting environments, reusable AI services that multiple business units can draw from without rebuilding from scratch, and governance frameworks that give the organization confidence to scale. When that investment is shared across downstream projects, the economics transform. Projects that were individually infeasible become viable. The total return across the portfolio far exceeds what any of those projects could have delivered independently. Beyond enabling other projects, Foundational investment also plays a critical risk management role. An organization that builds strong shared capabilities first is in a far stronger position to execute a Big Bet responsibly than one that attempts the same bet on a fragile data and infrastructure foundation.
The right AI portfolio composition looks different for every organization, and that is by design. A company facing aggressive competitive disruption may need to weight its portfolio toward Big Bets and Strategic investments, accepting higher risk in exchange for the potential to reshape its market position. A business in a stable industry focused on operational performance may lean more heavily toward Quick Wins and Low Risk initiatives that deliver steady, compounding returns. An organization early in its AI journey may need to prioritize Foundational investment before anything else, because without the right infrastructure and data foundations, every other archetype is harder to execute and more likely to stall.
What no organization should do is concentrate everything in one archetype. A portfolio of nothing but Low Risk projects may be safe, but it is unlikely to create meaningful competitive advantage or enterprise-level impact. A portfolio of nothing but Big Bets may be ambitious, but the risk of failure is high and there is nothing in the mix to maintain organizational momentum and credibility while the longer-horizon bets develop. The archetypes are not a menu to choose from. They are ingredients to be combined deliberately, in proportions that reflect where the organization is today, where it needs to go, and what it can realistically execute.
That calibration, deciding the right mix of archetypes for a specific organization at a specific moment, is precisely what an AI project prioritization methodology makes possible. It is the difference between a portfolio assembled by intention and one that simply accumulated over time.
AI Project Prioritization: The Methodology That Makes It Real
Recognizing the value of portfolio thinking is one thing. Building a portfolio in a principled, repeatable, and defensible way is another, especially in organizations where project advocacy is political and the loudest voice in the room has historically had the most influence over what gets funded.
This is where AI project prioritization comes in. Not prioritization in the informal sense of ranking a list, but prioritization as a structured methodology: a consistent, evidence-based process for evaluating every project candidate against the same criteria, surfacing the ones that belong in the portfolio, assigning them to the right archetypes, and assembling a mix that is coherent and defensible.
Effective AI project prioritization has to do several things simultaneously. It has to assess business value, but not business value alone, because a project with a compelling ROI that depends on data infrastructure that does not yet exist is not actually ready to execute. It has to account for organizational readiness, the people, process, and change management conditions that determine whether AI delivers its projected value or stalls in deployment. It has to surface synergies, identifying which projects share foundational dependencies and which become viable together that were not viable apart. And it has to produce not a single ranked list, but multiple views of the project landscape that reflect different portfolio lenses.
When AI project prioritization is done well, the portfolio assembles itself from the evidence rather than from the loudest voices. Resource conflicts become visible before they cause damage. Synergies get captured rather than missed. And the resulting portfolio becomes something the leadership team can present to the board not as a list of technology initiatives, but as a structured capital allocation decision with a clear strategic rationale, one that can be communicated, defended, and revisited on a regular cadence as priorities evolve.
In Part 2 of this series, we go deeper into the how: the SoT AI Portfolio Prioritization Framework, the scoring methodology, and the specific criteria and dimensions that make it rigorous and repeatable.
Is Your AI Project Prioritization Process Working?
Before reading Part 2, consider these questions honestly:
- When a new AI project is proposed, do you apply a consistent evaluation framework, or does assessment depend on who is in the room?
- Can you clearly connect your current AI investments to measurable business outcomes at the enterprise level, not just at the pilot or project level?
- Can you see across your entire AI initiative landscape clearly enough to identify where projects are competing for the same constrained resources?
- Have you ever greenlit a project only to find it blocked the progress of something more valuable that came after it?
- Are there projects in your current pipeline that look infeasible individually, but might become viable if evaluated together because they share infrastructure, data, or delivery resources?
- Does your portfolio include a deliberate Foundational layer, or is shared infrastructure consistently losing budget to projects with more visible near-term returns?
- Is your AI pipeline reviewed as a portfolio on a regular cadence, or only when a new project forces the conversation?
If most of these questions are difficult to answer, or if the answers vary depending on who in your organization you ask, your AI project prioritization process is likely overdue for a structured rethink.
The Shift That Changes Everything
The organizations that will lead in AI are not the ones with the most projects. They are the ones that have stopped treating projects as independent decisions and started managing them as a coherent AI portfolio, assembled with intention, balanced across archetypes, and constructed through an AI project prioritization process that is consistent and repeatable.
That is how AI activity becomes enterprise impact. Not through any single initiative, however promising. Through the compounding effect of the right mix, built through the right methodology, managed as a strategic capability rather than a collection of one-off bets.
The shift from one-off thinking to portfolio methodology is not a technology decision. It is an operating model decision, and one your organization can make.
Part 2 of this series introduces the SoT AI Portfolio Prioritization Framework, the four-step methodology, scoring criteria, and AI dimensions that make portfolio construction rigorous, repeatable, and defensible.
This article is part of a continuing series aimed at providing senior leaders and managers with a practical working knowledge of artificial intelligence and how to manage it as a business capability.
Thanks for reading this post. If you found this post useful, please share it with your network. Please subscribe to our newsletter and be notified of new blog articles we will be posting. You can also follow us on Twitter (@strategythings), LinkedIn or Facebook.
Related posts:
How the Right AI Model Translates Into Decisions, Strategy, and Results.
The AI Build–Buy–Partner Decision: A Strategic Framework for Executives
AI Is Everywhere. Enterprise Impact Isn’t. Here’s the Structure That Closes the Gap.