For decades, the process of software development has resembled a relay race. Business analysts pass requirements to product managers, who hand specifications to developers, who throw code over the wall to QA, who eventually ship to operations teams. Each handoff introduces delays, miscommunication and lost details.Amanda Silver, corporate vice president of product in Microsoft’s Developer Division, told The New Stack she has watched this relay race across thousands of teams. But over the past year, she’s witnessed something unprecedented. This entire race has evolved into unified workflows powered by AI agents.“Software development really was this long, slow relay with multiple handoffs throughout development phases, from ideation to spec-ing [establishing specifications] to prototyping to testing and deployment,” Silver said. “Now we’re seeing all of this collapsed in our tools — you can go through multistep, complex coding tasks from writing to testing to debugging, all from a single prompt.”From Pair to Peer ProgrammingThe transformation begins with a shift in how AI assists developers. GitHub Copilot started as a “pair programmer” — an AI assistant looking over your shoulder. The latest generation represents something different: peer programmers you can assign complete tasks to.“We’ve introduced agent mode, the Coding Agent, taking GitHub Copilot from a pair programmer to a peer programmer you can assign tasks to,” Silver told The New Stack. “It handles bug fixes, writes tests or implements full features.”Instead of line-by-line suggestions, developers now delegate entire workflows — from bug investigation and fixes with detailed pull requests to complete test suite generation including edge cases, end-to-end feature implementation from database to UI and automatic documentation with deployment guides.Collapsing the Entire Life CycleThe real breakthrough occurs when AI agents operate across the complete software development life cycle — what Silver calls “agentic DevOps.”Agents can now analyze business requirements and suggest technical approaches during planning, while tools like GitHub Spark and GitHub Copilot let developers describe applications in natural language and see them come alive instantly during prototyping. During development, coding agents handle implementation plus comprehensive testing, security scanning and performance optimization. In operations, SRE agents for Azure monitor systems 24/7, troubleshoot autonomously and resolve many incidents without human intervention, Silver said.This integration transforms “multistep, complex coding tasks” into unified workflows rather than discrete handoffs between specialized teams.The Economics of ExperimentationMoreover, Silver said she believes agentic DevOps’ most significant impact will be economic. By reducing costs across every development phase, AI agents could enable a new wave of software experimentation similar to cloud computing’s infrastructure revolution.“This will dramatically reduce total cost of ownership and experimentation costs, whether introducing a brand new solution — starting a company with an idea you want to prototype — or building solutions within organizations,” Silver explained.Before cloud computing, starting a software company required significant upfront infrastructure investment. Cloud platforms reduced those barriers to nearly zero, enabling the startup explosion of the past two decades.Agentic DevOps could similarly transform development itself by enabling prototype development in hours rather than weeks, rapid feature experimentation for market validation, automated legacy modernization that removes innovation barriers and automated operations that reduce complexity overhead, Silver said.Real-World ImpactEarly adopters are already seeing dramatic results. Silver cites Carvana using the “GitHub Copilot coding agent to move from idea to specification to production in minutes, increasing velocity and channeling energy toward higher-level creative work.”And Cathay “rolled out GitHub Copilot to over 1,000 developers, streamlining delivery, improving security and developer experience so they can focus on innovation.”Transforming Developer JoyBeyond productivity, agentic DevOps changes what it means to be a software developer. Silver emphasized that the goal isn’t just speed but joy in development.“Integrating AI isn’t just about acceleration — it’s about building better, more secure software that’s more of a joy to build,” she explained. “This allows developers to offload tedious, repetitive, mundane tasks and headaches they don’t like.”Specific improvements include the elimination of drudgery through automated dependency updates, boilerplate generation and basic debugging, allowing creative focus on architecture, UX design and problem-solving. Developers also experience reduced on-call burden with automated incident response meaning fewer 2 a.m. wakeups, while improved quality comes from consistent patterns, security scanning and optimization suggestions.“Nobody likes being woken up at 2 a.m. for live site incidents,” Silver noted. “Agentic DevOps lets developers get better sleep and bring their creative brain to work for innovative problem-solving.”Industry RealityThis transformation is happening hastily. GitHub Copilot has crossed 20 million monthly active users among GitHub’s 150 million developers. Gartner predicts 33% of enterprise applications will include agentic AI by 2028, with 80% of enterprises having AI-enabled apps in production by 2026.These numbers indicate that agentic DevOps is not simply emerging; it’s becoming standard.The End of the Relay RaceSilver’s vision represents more than tool evolution — it’s a reimagining of software development. By eliminating handoffs and delays that have been part of software development for decades, AI agents enable ideas to flow directly into production with minimal human friction.“The nature of software engineering is dramatically changing,” Silver concluded. “We’re tackling the most miserable, soul-draining parts of the job, transforming them so developers can focus on creative aspects they really enjoy.”The post Microsoft: AI Agents Are Winning the DevOps Relay Race appeared first on The New Stack.