McKinsey will advise clients on which engineering projects to pursue and how to redesign teams and workflows around AI. Cognition will supply Devin, its coding agent, which the company says can understand, plan, write and test code.
Devin’s launch in early 2024 unsettled some software engineers, who worried AI could reshape their work or threaten their livelihoods. Developers’ jobs have since shifted across industries, from writing code toward guiding AI output. Lari Hämäläinen, a senior partner in McKinsey’s technology, media and telecommunications practice, said many tools were still best suited to narrow tasks such as finding technical guidance or syntax details until the final months of 2025. More capable models have since enabled longer work to be handed off to AI, he said.
The pace of technical change can outstrip companies’ ability to redesign their processes, Hämäläinen said. He called the shift a “step-change phenomenon” that “requires a complete rethink” of software and product development. Adoption can also produce uneven results: in a recent McKinsey survey of more than 300 product and engineering leaders, the top quarter of organizations reported productivity gains of more than double, while nearly 30% said productivity had fallen. Scaling a tool from a small trial to thousands of developers is a separate challenge, he said.
Managing the cost of AI work
Companies including Goldman Sachs, Cisco and JPMorgan have become more selective about who can use AI tools and how, as they look for returns and ways to control costs. Art Levy, Cognition’s vice president of global partnerships, said Devin can draw on multiple AI models, allowing companies to use more expensive frontier models for difficult engineering and cheaper ones for other tasks. He compared the choice to not needing “a Ferrari to drive to the grocery store.”
Levy said the approach could help with delayed software-modernization work. At one client, he said, an enterprise resource planning upgrade expected to take five years is being compressed into 18 months. The question, he said, is no longer only whether the work can be done, but whether it can be done cheaply.

Comments (0)
No comments yet. Be the first!