Your AI Maturity Score Is Probably Wrong
Your organization has AI coding tools. Your developers use them every day. You may have an AI strategy, a governance program, and a growing list of AI projects. Does this make your organization AI mature? Most organizations measure what is easy to count: licenses, adoption rates, generated code, and reported hours saved. These metrics show that developers use AI. They do not show that the organization builds better software with AI. This talk compares the major AI maturity models from CMU/SEI, DORA, NIST, and OWASP. We will show what each model measures, where the models overlap, and where they fail. Then we will discuss a practical framework that answers three questions: Adoption: Do we use AI? Capability: Do we have the engineering practices to use AI well? Outcomes: Can we prove that AI improves delivery speed, quality, security, or developer experience? You will see why high adoption often hides low maturity, why individual productivity metrics mislead leaders, and what to measure instead. The goal isn’t another five-level AI maturity staircase. It’s a way to gather evidence that an organization is actually getting better at building software in an AI-enabled world.


