Moving away from Agile: What’s Next – Martin Harrysson & Natasha Maniar, McKinsey & Company

Martin Harrysson and Natasha Maniar from McKinsey’s Software X practice present findings from a survey of approximately 300 enterprises on AI adoption in software development—and explain why most organizations are only seeing 5–15% productivity improvements despite the availability of powerful AI tools. Their diagnosis: the bottleneck is not the technology but the operating model. Team structures, work allocation processes, and collaboration patterns designed for Agile haven’t been updated to match the new pace of AI-assisted development.

Drawing on a client engagement with a major international bank, they outline specific interventions: agents assigned sprint stories based on team velocity data, co-created acceptance criteria with AI to reduce downstream rework, squads reorganized by workflow type rather than function, and agents monitoring cross-repository impacts in the background. They also describe the shift to spec-driven development—where product managers iterate on detailed specs with agents rather than writing long PRDs—and smaller pods of three to five full-stack product builders replacing the traditional two-pizza team. The talk includes an analysis of how Cursor operates internally as an AI-native organization.

Organizations that have made these structural shifts are seeing five to six times faster time-to-market alongside higher artifact consistency. The McKinsey researchers position this as the next Agile-scale transformation: not a tool upgrade, but a fundamental change in how software teams are organized and how work flows through them.

📺 Source: AI Engineer · Published December 12, 2025
🏷️ Format: Deep Dive