Ninety-five percent. That’s the number CMU’s Software Engineering Institute and Accenture report in their joint AI Adoption Maturity Model publication, according to the research itself: 95% of surveyed organizations realize zero returns on their AI investments. The companion figure, only 8% of enterprises successfully embed AI into core business strategy, frames the same gap from the other side.
Read those numbers with one fact in mind: Accenture co-authored the maturity model framework it’s now recommending organizations adopt, and Accenture sells the consulting services to implement it. That’s not disqualifying, CMU’s SEI is a credible T1 research institution, and joint industry-academic frameworks are standard practice. But the 95% and 8% figures come from this specific joint release, not from an independent third-party survey. Attribution matters here. Per the CMU SEI Digital Library publication, these figures reflect the survey methodology designed for this research initiative.
The framework itself structures AI adoption across eight organizational dimensions. The report names these dimensions, the Builder’s production note from the Filter flags that dimension names should be verified against the CMU SEI Digital Library before publication; the source document is T1 but the specific URL wasn’t provided in the source package. What’s confirmed is the eight-dimension structure and the framework’s focus on organizational readiness across engineering, strategy, governance, and culture categories, per the joint Accenture press release.
Disputed Claim
The framework’s core critique, that organizations are treating AI as a technology implementation problem rather than an organizational transformation problem, isn’t new. What’s useful about the CMU SEI framing is the SEI’s credibility in software engineering maturity models specifically. The SEI built the Capability Maturity Model Integration (CMMI) framework, which became the standard for software process improvement in government and defense contracting. If the AI Adoption Maturity Model borrows from that methodological tradition, it’s likely structured around maturity levels (initial, repeatable, defined, managed, optimizing) applied to each organizational dimension, though the specific maturity stage architecture should be verified against the primary document.
The part nobody mentions in maturity model releases: most organizations already know they’re not capturing AI value. They don’t need another framework confirming the problem. What they need is a framework that identifies the specific bottleneck, whether it’s data infrastructure, governance gaps, skill deficits, or leadership misalignment, and prioritizes which dimension to address first given their actual organizational constraints. Whether the CMU/Accenture model does this, or whether it provides a comprehensive but non-prioritized list of everything that needs to improve, is the question practitioners should ask before adopting it as a diagnostic tool.
Analysis
CMU SEI's CMMI framework became a standard in government and defense software contracting. If this AI maturity model follows that tradition, it will carry institutional weight in procurement and government AI adoption conversations, regardless of the Accenture co-authorship. Watch whether DoD or civilian agencies cite it in AI acquisition guidance over the next 12 months.
Don’t treat the 95%/8% figures as settled industry benchmarks. They’re findings from a single joint publication with a specific survey methodology. They may be directionally accurate, the AI value realization gap is well-documented across multiple independent sources, but the specific numbers belong to this report.
Verify the eight dimension names against the CMU SEI Digital Library before using them in internal assessments or board presentations. The framework has credible institutional backing. The specific methodology deserves the same scrutiny you’d apply to any consulting-affiliated research before building strategy around it.