What Arthur Miller's AI-SDLC Maturity Model Means for Companies Making the Shift to Spec-Driven Development
SCOTTSDALE, AZ / October 22, 2026 / Every enterprise is somewhere on the road to AI-driven software delivery, whether it has chosen a position or not. Arthur Miller's contribution is a map of the road itself.
Miller is the enterprise architect behind Zelle and Real-Time Payments infrastructure, the founder and CEO of financial platform company FiQIS, and the author of the new book AI-SDLC and the New Enterprise Architecture Approach. The book arrives as AI-native delivery reports 97 to 98 percent calendar compression in early enterprise deployments, alongside Miller's caution, earned across payments, defense and clinical software, that a 97 percent faster path to the wrong system is a 97 percent faster mistake.
What separates his framework from the industry's usual advice is that it doesn't ask companies to start over. It asks them to look at what they already produce.
Every organization is full of requirements documents, architecture plans, project schedules and release notes. Today those artifacts are written for humans to read once and file away. Miller's model converts them, one at a time, into living specifications that AI systems can execute against and be checked against. The transformation happens artifact by artifact, not through a company-wide program, which is what makes it survivable for a real business with real deadlines.
The maturity model gives leaders three recognizable stages.
At the Guided level, specifications steer collaboration between people and AI. Humans direct the work, machines accelerate it, and the spec keeps everyone aligned on intent. Most organizations experimenting seriously with AI development are reaching for this stage now, whether they use the vocabulary or not.
At the Contracted level, the specification hardens into a binding interface. It becomes precise enough that outputs can be automatically verified against it, which is the point where speed stops depending on trust in the machine and starts depending on the quality of the contract.
At the Generated level, the specification becomes the source of truth from which working systems are produced. Humans govern the spec rather than the code, and judgment moves to the highest point of leverage in the entire lifecycle.
Wrapped around the maturity path is the governance that Miller, a veteran of regulated industries, refuses to treat as optional. His eight-stage lifecycle, from Discovery through Analysis, assigns every stage a tollgate with a named human owner. Agents produce, humans judge, and someone always answers for what ships.