Microsoft said on 15 October that it has launched PRAISE — Preceptorship for AI-accelerated Software Engineering — an internal programme that pairs early-career engineers with senior preceptors, a model it says is borrowed from nursing. Preceptors coach juniors on using AI tools while building systems thinking and judgement, and the company is working with Carnegie Mellon University on a curriculum taught alongside the on-the-job training, including small-group discussions meant to connect computer-science theory to production work. The post is signed by Mark Russinovich, CTO and deputy CISO of Microsoft Azure, and Scott Hanselman. Michael Hilton, a teaching professor in Carnegie Mellon's Software and Societal Systems Department, is quoted saying that combining academic and practitioner expertise produces training that stands up to daily development realities. Microsoft's argument is that AI changes the work but not human responsibility for the outcome.
Our takeOur take: The load-bearing claim is that review, not code production, is now the constraint — testable, yet the post offers no measure of what preceptees ship or how often their output is rejected. The structure is concrete: a named senior preceptor, a Carnegie Mellon curriculum run alongside production work, and small-group sessions tying theory to tickets. That matters technically because AI-assisted work shifts effort from writing to specifying and verifying, skills learned by reviewing someone else's diff. Treat the responsibility claim as a design principle until there is data.