Managing Your AI Portfolio: Prioritize What Matters

Every executive I talk to has the same problem: too many AI ideas and not enough of anything else - budget, talent, attention, time. Without a framework to cut through the noise, resources get spread thin and strategic opportunities get missed while you’re chasing the ones that were easiest to get started.

Here’s what I’d do.

The real problem isn’t picking the wrong ideas

It’s not having a consistent way to compare ideas across different departments with different goals and different baselines. When a finance team proposes an AI model for cash flow forecasting and a marketing team proposes an AI-powered campaign personalisation engine, how do you compare them fairly? You can’t, without a common scoring framework.

The absence of that framework is why most companies end up stuck in basic automation - chasing the easy tasks rather than the transformational projects that actually change how the business competes.

Score every project on four dimensions

A decision matrix that evaluates each potential initiative against four criteria gives you the structure to make defensible choices.

Business impact first. Estimate revenue growth, cost savings, risk reduction, or strategic advantage. Projects with a direct line to EBITDA improvement or competitive differentiation rank highest. If you can’t quantify the business case, that’s information - it means either the case isn’t there or you don’t understand the opportunity well enough to pursue it.

Feasibility second. Assess data quality, technology readiness, talent availability, and realistic timelines. Projects requiring extensive data cleaning or rare expertise score lower unless their impact is exceptional. Be honest here - wishful feasibility is how budgets get blown.

Risk third. Evaluate regulatory, ethical, and operational exposure. High-risk projects require heavier governance and longer approval cycles. Weigh whether the potential rewards justify the added complexity before you commit.

Strategic fit fourth. Does this project align with your core strengths and long-term goals? Does it amplify your unique competitive advantage - or is it a generic tool any competitor could equally adopt? If it’s the latter, it’s table stakes at best, not a source of differentiation.

Use a 2x2 to force the conversation

Plot projects on a grid of impact versus feasibility. Prioritise the high-impact, high-feasibility quadrant first. Re-evaluate everything else quarterly as data quality, regulations, and market conditions shift.

In your portfolio review, maintain a mix of quick wins and longer-term transformational projects. Quick wins sustain momentum and build the organisational credibility that larger bets require. Capability-building initiatives create the foundation for them.

Say no to low-impact projects, even when they’re easy

This is the discipline that separates well-run AI portfolios from sprawling ones. Easy doesn’t mean valuable. A project that’s simple to execute but doesn’t move any needle worth mentioning is still consuming resources that could go somewhere that matters.

Quantifying impact is never perfectly clean - use ranges rather than single-point estimates and include qualitative benefits like brand differentiation or talent retention. Be prepared to advance a project even with medium feasibility when strategic fit is compelling. Sometimes building the capability is worth the extra effort.

But the flip side: when a project scores low on impact, say no. Clearly, and without ambiguity. “We might get to this later” usually means “we’re keeping this on the list to avoid a difficult conversation.” Have the conversation instead.

A disciplined portfolio management process ensures that resources flow to the right initiatives. Score projects consistently, revisit those decisions as new information emerges, and stay honest about what’s generating value versus what’s generating activity.