If your strategic AI plan assumes responsible AI is now operationally mature in most organisations, the 2025 data shows you are planning ahead of where the field actually is. McKinsey's RAI maturity score improved across all regions from 2024 to 2025, but the global average is still 2.3 on a four-point scale. Most organisations are still in the integration phase, not the fully-operational phase.

Responsible AI maturity by region, 2024 vs. 2025 (McKinsey 4-point scale):

The maturity framework: McKinsey measures RAI maturity on a four-point scale. Level 1 (Foundational): RAI practices have been developed. Level 2 (Integrating): those practices are being integrated into the organisation. Level 3 (Fully in place): all necessary practices are operational. Level 4 (Comprehensive and proactive): RAI practices are fully operational with proactive improvement.

The 2024-to-2025 movement: the global average rose from 2.0 to 2.3, a 0.3-point improvement in twelve months. Asia-Pacific (excluding China, including India) leads at 2.5 (up from 2.2). Europe rose from 2.0 to 2.3. Latin America showed the largest year-over-year gain, climbing from 1.8 to 2.2. North America registered the smallest improvement, moving from 2.1 to 2.2.

The picture across regions: meaningful improvement everywhere, but the global average is firmly in the Integrating phase. No region has reached the Fully in place phase (level 3). The most mature region, Asia-Pacific at 2.5, is halfway between Integrating and Fully in place. Most organisations are still building infrastructure, not operating it.

This is a substantially different baseline than the strategic AI narrative often assumes. Plans built around "RAI is mature; the question is which specific framework to adopt" are operating against data that shows most organisations are still building foundational and integration-stage capacity. The procurement and partnership conversations of 2025–2026 will frequently encounter organisations operating below their stated maturity level, because the maturity scale itself is generous, and because integration takes longer than adoption.

Three observations follow from the regional pattern.

The first: the regions with the largest 2024-to-2025 gains are the regions starting from lower baselines. Latin America's 1.8-to-2.2 gain represents organisations crossing from Foundational into Integrating. Asia-Pacific's 2.2-to-2.5 gain represents organisations advancing within Integrating. The improvement curve is steeper at the lower end of the maturity scale and flatter at the higher end. Organisations near the top of the scale find each additional 0.1 of maturity harder to capture than organisations near the bottom.

The second: North America's small improvement (2.1 to 2.2) is the divergent data point. The region with the largest cumulative AI investment and the most established AI vendor ecosystem registered the smallest year-over-year RAI maturity gain. The reasons are not directly visible in the data, but candidates include: organisational complexity slowing integration in larger US enterprises, regulatory environment uncertainty (the Biden executive order revocation in early 2025), or simply that North American organisations were already higher in 2024 and have less low-hanging improvement available.

The third: the survey explicitly excludes responses from China, which limits the geographic scope. The 2025 maturity picture is missing the AI ecosystem that is now leading in RAI research output (see #62 China overtook the US in RAI research). Strategic planning that assumes the McKinsey maturity numbers represent the global picture is missing data about the world's second-largest AI economy.

For strategic AI planning in 2026, three implications follow.

The first: maturity-stage targets should be set explicitly and tracked over time. A plan that says "the organisation will reach Fully in place (level 3) within 18 months" is a measurable strategic commitment. A plan that says "the organisation will improve responsible AI practices" is a directional statement that cannot be tracked. The McKinsey scale provides the structure for measurement; strategic plans should adopt it explicitly.

The second: the gap between current state and target state should drive investment sizing. An organisation at level 2.0 targeting level 3.0 within 24 months needs to plan substantial infrastructure build-out: hiring, tooling, governance structure, policy development. An organisation at level 2.5 targeting the same level 3.0 needs different investment, focused on consolidating and operationalising practices already in place. The investment plan should be sized to the gap, not to a uniform RAI budget across all organisations.

The third: the maturity ceiling that organisations should plan for is meaningful, not perfect. Level 4 (Comprehensive and proactive) is the top of the scale, but no region averages above 2.5. Achieving level 4 is rare. The realistic target for most organisations over the next 24–36 months is level 3 (Fully in place), and the strategic plan should reflect that ambition without overclaiming.

The trajectory: RAI maturity will continue improving in 2026–2027. The slope depends on regulatory pressure (EU AI Act phasing in is a strong driver), investment maturity (large enterprises now spend $25M–$50M on RAI operationalisation, see #59), and the integration capacity of the organisations themselves. The most defensible planning anchor is that the global average rises from 2.3 toward 2.8–3.0 over the next 24 months, with the leading regions (Asia-Pacific) crossing into Fully in place territory first and others following.

For executives setting AI direction, the planning anchor is "responsible AI is mid-implementation across the global enterprise population, with significant integration work remaining at most organisations including peers and partners." By 2027, the leading regions will likely cross into Fully in place territory; lagging regions will still be operationalising. Plans built on this anchor will be aligned with where the data shows the field is. Plans built on the assumption that RAI is operationally mature will be ahead of the field in language but behind it in execution.

Sources

  • Primary: Stanford AI Index 2026, Chapter 3 (Responsible AI) 3.3 — hai.stanford.edu/ai-index/2026
  • Survey data: McKinsey & Company "State of AI" Survey, 2025 — RAI maturity 4-point scale by region (survey excludes China)
  • Regional RAI research context: cross-reference China overtook the US in RAI research in one year