SEI Releases AI Engineering Practices to Accelerate Mission-Capable AI

April 23, 2026—The Software Engineering Institute (SEI) yesterday released updated recommendations for organizations building, acquiring, and integrating artificial intelligence (AI) capabilities into business and mission systems. AI Engineering: 12 Foundational Practices, a white paper for decision makers, describes the disciplined groundwork for turning AI capability into mission value.

Engineering for Speed and Sustained Value

The paper’s practices aim to accelerate AI adoption in defense and national security organizations, in line with recent pushes from the National Institute of Standards and Technology (NIST), the White House, and the Department of War. By asking critical engineering questions in the early stages of building and adopting AI technologies and capabilities, executives, program managers, and technical leadership forestall problems and hasten mission value.

“Sound engineering is what makes rapid deployment possible,” the SEI’s Matt Gaston, Eric Heim, and Hollen Barmer write in the paper. “Without the AI Engineering practices described here, speed creates risk rather than advantage.”

The paper covers practices such as choosing the right problems for AI to solve, selecting the appropriate solutions, and managing change in AI technology, among other practices.

Evaluation of AI Systems Elevated

Yesterday’s release updates the original 11 practices the SEI published in 2019. A new practice recommends investing in evaluation as a first-class engineering challenge. The practice accounts for the broadening adoption of AI and agentic systems in the defense domain.

Other updates to the practices reflect the significant, continuing evolution of building, using, and deploying AI. Today, AI capabilities arise from training bespoke models, consuming foundation models via an application programming interface, fine tuning open-weight models, and assembling compound systems of AI and non-AI components. The paper also revisits security, traceability, uncertainty, oversight, modularity, and lifecycle data considerations in the context of foundation models and agentic systems.

“AI capabilities have advanced substantially along with the engineering challenges,” write the paper’s authors. “Sound AI Engineering is what separates organizations delivering sustained value from those that cannot move beyond prototypes.”

Discipline for Mission

The 12 practices distill the SEI’s experience assisting government organizations, observing the practices of teams developing and implementing AI capabilities, and learning lessons from industry, as well as the SEI’s legacy with software engineering as a discipline. Drawing on this experience, the SEI is developing AI Engineering to help the defense and national security communities develop, deploy, operate, and evolve mission capabilities that leverage rapidly evolving AI and machine learning technologies. Future SEI publications will detail the components of AI Engineering.

Download AI Engineering: 12 Foundational Practices from the SEI Digital Library. Learn more about AI Engineering on our website.