No consulting firm uses a single matrix

Every time AI prioritization is discussed, someone asks which is "the best matrix".

It is the wrong question. And the firms that manage the most strategic portfolios in the world prove it: none of them uses just one.

• McKinsey combines Three Horizons + GE/McKinsey Matrix + bubble charts + financial analysis.
• BCG starts from the BCG Matrix and adds value creation, scenarios and competitive capabilities.
• Gartner works from capabilities, organizational maturity, risk and business value.
• Deloitte and Accenture open with Impact × Feasibility and then add capabilities, governance, architecture and roadmap.

Different methodologies, one common principle:

Strategic decisions are not made with a single matrix.

An excellent first filter

The Impact × Feasibility matrix is an excellent first filter. Four quadrants, fast decision:

• I. High impact + high feasibility → execute first
• II. High impact + low feasibility → strategic bets, prepare
• III. Low impact + high feasibility → incremental improvements, useful for learning
• IV. Low impact + low feasibility → discard or rethink

The problem appears when that filter is used as a complete decision model. Because there are things two axes cannot see:

• Regulatory and ethical risk
• Data quality and availability
• Reuse of existing AI assets
• Required organizational capabilities
• Dependencies between initiatives
• Time horizon (H1 / H2 / H3)
• Technical debt and scalability

A concrete example of the blind spot: a medium-feasibility initiative that installs a feature store reusable by eight other initiatives is worth far more than its score in the matrix. The matrix penalizes it. The portfolio should reward it.

Six frameworks against the dimensions of AI

In the carousel that accompanied this series I reviewed the six most used prioritization frameworks (Impact × Feasibility, Impact × Effort, RICE, WSJF, Three Horizons and GE/McKinsey) against the dimensions an AI initiative really demands. The result is more uncomfortable than I expected: each framework answers part of the problem. None answers the whole problem. None was designed for a world with data, models, agents and algorithmic governance.

AI introduced dimensions that did not exist when these frameworks were conceived: data, models, agents, algorithmic governance, value reinvestment.

The natural evolution is not to look for a new matrix to replace the previous ones. It is to build AI Portfolio Decision Frameworks that integrate them as layers of the same decision process. The pillar article of the series develops that full framework.

Which criteria does your organization use to prioritize its AI investments?