Martin Harrysson is a Bay Area–based senior partner at McKinsey & Company and a leader in its SoftwareX practice, where he helps technology, software, and financial-services companies rethink how they build products. His central concern is whether AI can improve the entire software organization, not simply make individual programmers write code faster.
Harrysson studied computer science and engineering at MIT and began his career as a software engineer, experiencing firsthand the organizational upheaval accompanying a company’s transition to Agile. At McKinsey, he expanded into product strategy, engineering effectiveness, and operating-model design, helping clients reshape software R&D organizations, develop cloud-product strategies, strengthen product management, and redesign fintech technology operations.
AI-native software operating models: Faster code generation can create bottlenecks in manual review, security, planning, and coordination. Harrysson supports redesigning workflows across the product development life cycle, with continuous planning and engineering roles adapted to collaboration with agents. His research into AI-enabled software development examined nearly 300 publicly traded companies.
Specification-driven development: Agents work more reliably when teams provide clear requirements, architectural context, explicit acceptance criteria, and defined security expectations. Better specifications reduce plausible-looking but misdirected output and downstream rework.
Smaller multidisciplinary engineering pods: Compact teams with broader product and technical fluency can replace rigid handoffs among engineering, product management, design, and quality assurance. Engineers increasingly direct agents, evaluate their work, and make architectural decisions.
Outcome-based engineering measurement: Tool adoption and generated code are weak substitutes for delivery speed, software quality, resilience, developer experience, customer outcomes, and economic impact. Enterprise adoption also requires practical coaching, revised incentives, updated responsibilities, and sustained organizational change.
Faster coding exposes slower coordination, review and planning. Capturing the gains requires changing how teams allocate work, define acceptance criteria and measure outcomes.
