McKinsey

Martin Harrison and Natasha Maniar presented McKinsey's case against Agile on 23 November 2025: despite spectacular individual productivity gains from coding agents, many large enterprises plateau at 5 to 15% overall improvement. Their explanation is a set of new bottlenecks, manual code review that cannot keep pace with generation, human collaboration built for a slower rhythm, and rising technical complexity. What McKinsey observed among "Top Performers" is a shift to AI-native workflows: user stories replaced by specs that product managers and developers iterate on with agents, Agile teams of eight to ten broken into pods of three to five with consolidated full-stack roles, and quarterly planning compressed into continuous cycles. One international bank reorganized its teams around bug fixing versus greenfield workflows and increased code merges by 51%. Harrison and Maniar put the constraint on change management: 70% of companies have not adapted their job descriptions.

Elsewhere, McKinsey shows up as a supplier of numbers. Alexandre Frizzo reaches for the McKinsey figure of 500% productivity in flow state to account for what a year with Claude Code cost his attention. Max Thilén cites McKinsey's $3-5 trillion projection for global agentic commerce revenue by 2030. Michael Nuñez's report on Mistral AI's European compute build-out uses McKinsey's $5.2 trillion estimate of datacenter capex needed by 2030. The Batch relays McKinsey's projection of 5 million humanoids in factories by 2040, against roughly 200 in 2026. The firm's reading of the agent era travels mostly as figures other people quote.