Microsoft and Carnegie Mellon launch PRAISE to train junior engineers for AI coding

What Microsoft and Carnegie Mellon announced

Microsoft and Carnegie Mellon University have started a program to train early-career software engineers for a job in which AI writes much of the code. In a Command Line post on 15 October 2026, Microsoft said it had launched PRAISE, short for Preceptorship for AI-accelerated Software Engineering, an internal program that pairs junior engineers with senior engineering "preceptors". Carnegie Mellon's School of Computer Science published its own announcement the same day, describing an online certificate that the Microsoft participants take alongside their full-time jobs.

The pitch is a direct answer to a question many new graduates are asking: if AI can produce working code in seconds, how does a beginner ever learn to engineer software? Microsoft's answer is that the career is not going away. "Of course we need junior software engineers," the post says. "Senior engineers ensure the technology we ship works safely and effectively, and today's junior engineers will be tomorrow's senior engineers."

The company frames the problem as one of judgement rather than typing speed. It asks whether people who doubt the value of a computer science degree would "ship AI code in a finished product sight unseen", and adds, "please, we implore you, say no." Its summary line is that "AI changes the work; it does not remove human responsibility for outcomes."

How the preceptorship works

The idea is borrowed from nursing, where newly qualified nurses work under an experienced nurse until they can practise confidently on their own. In PRAISE, a junior engineer, called a preceptee, is paired with a senior engineer on real Microsoft work. The preceptor explains how they approach a problem and why, shares their sense of what good code and design look like, and advises the preceptee on where to focus based on their skills.

Microsoft says the goal is not to make newcomers code as though AI never arrived. Preceptors coach juniors to use AI tools "thoughtfully, while building strong systems thinking, taste, and judgement". Junior engineers are taught to test assumptions and to talk through unfamiliar changes with colleagues, so that producing a fix becomes a chance to learn rather than just a closed ticket. The post also asks senior engineers to accept that juniors may lack the confidence to contribute at full speed, and to give them room to learn by doing rather than by watching someone else solve the problem.

Microsoft compares the shift to the arrival of compilers, which took away some low-level work without removing the need to understand what the machine is doing. In its words, more engineering time now goes on design and architecture than on writing every line, but "when a car breaks down, it takes someone who knows what's going on under the hood to get it back on the road."

The Carnegie Mellon courses

The academic half of PRAISE is a certificate run by Carnegie Mellon's Kenneth C. Griffin School of Computer Science. Microsoft employees take online courses while keeping their jobs. Carnegie Mellon's announcement profiles Jace Watson, a Microsoft software engineer in Seattle, who said AI tools were not a big part of his degree but became part of his daily work as soon as he joined Microsoft.

This autumn, participants are taking Machine Learning in Production and Preceptorship in AI-Augmented Software Engineering I, which Carnegie Mellon says cover building machine learning into software and using AI tools while developing it. In the spring they move on to DevOps: Engineering for Secure Development and Deployment and the second part of the preceptorship course. Carnegie Mellon says the credits can count towards its master's degree in software engineering.

The program is directed by Jonathan Aldrich, a professor in the university's Software and Societal Systems Department. "With so many companies worldwide faced with upheaval due to AI in software development, I believe this program will become a template for rapidly leveling up new developers across the industry," he said. Carnegie Mellon says it hopes to grow the certificate and perhaps open it to other industry partners, but for now it is a Microsoft-only program.

Teaching professor Michael Hilton, quoted by both organisations, summed up the gap the courses are meant to fill: "It used to be that early career software engineers would do a lot of tasks that would build up their expertise."

Why Microsoft says junior engineers are falling behind

Carnegie Mellon says the program grew out of an October 2025 visit to the school by Mark Russinovich, chief technology officer of Microsoft Azure and a Carnegie Mellon alumnus, who talked about how AI tools help senior engineers far more than junior ones. "AI changes the work, but it doesn't eliminate the need for junior engineers or for human judgment and accountability," he said in the announcement.

Russinovich and Scott Hanselman, Microsoft's vice president for developer community, set out the argument in detail in a Communications of the ACM opinion piece, Redefining the Software Engineering Profession for AI, posted on 5 March 2026. They write that coding agents give experienced engineers an "AI boost" while imposing an "AI drag" on early-career developers who lack the context to steer and check what the agents produce. The result, they say, is a new incentive: "Hire seniors, automate juniors."

Their examples are concrete. In Microsoft's Project Societas, the internal name for its Office Agent, seven part-time engineers built a consumer preview in ten weeks with more than 110,000 lines of code, 98% of it AI-generated. But they also show an agent that "fixed" a crash caused by a race condition by inserting a sleep, which hides the bug without solving it. An early-career developer, they argue, might accept that fix because the tests pass; only someone who understands the synchronisation code would spot the problem.

The essay also sketches what such a program could look like. It proposes preceptors drawn from a trained subset of senior engineers, each looking after three to five juniors, in programs lasting at least a year. It suggests coding assistants could offer an early-career mode that coaches with questions before writing code, and let preceptors review learners' chat logs to spot gaps. Neither announcement says whether Microsoft has built those tools yet.

How PRAISE fits the evidence on AI and junior workers

PRAISE arrives in the same week as fresh data on how AI is reshaping junior work in software. On 16 October Anthropic published an internal study finding that the junior share of its technical hires fell from 27% to 9% as agent hours overtook human hours six to one. Its own engineering team described layered AI review on the Claude Code codebase, where about 90% of code is written by Claude and humans step in mainly on high-risk changes.

Outside research points the same way on skills. A Google-funded randomised trial with 133 patent lawyers found that AI improved drafts on average but left juniors no more skilled once the tool was taken away, while seniors improved. And a Harvard study of 718 firms found that coding agents raised code output by up to 30% without a matching rise in finished work, because review became the bottleneck. That is exactly the work PRAISE is trying to teach juniors to do well.

Russinovich and Hanselman cite labour data showing that employment of 22 to 25-year-olds in highly AI-exposed jobs such as software development fell by roughly 13% after GPT-4's release, even as senior roles grew. Against that backdrop, two senior Microsoft executives arguing in public that companies should keep hiring juniors and invest in them, even though they "initially reduce capacity", is a notable position.

What PRAISE does not tell us yet

Neither Microsoft nor Carnegie Mellon has said how many engineers are enrolled, how they were chosen, how many preceptors are involved, or what the program costs. Microsoft describes PRAISE as internal, and there is no public application route for engineers at other companies. The ratios and tools in the ACM essay are proposals by two Microsoft executives, not published results from the program.

There are no outcome measures yet either. The program began this semester, so there is no data on whether preceptees learn faster, stay longer or ship fewer bugs than peers without a preceptor, and Microsoft has not said whether it will publish any. Training does not answer the hiring question on its own: the program helps engineers Microsoft has already hired, and says nothing about how many junior roles it plans to fill.

Still, PRAISE is one of the first named, structured responses from a major software company to a problem that research keeps flagging. We will update this article if Microsoft or Carnegie Mellon publish enrolment numbers, early results, or plans to open the certificate to other employers.