McKinsey and Cognition Team Up to Prove AI Coding Pays
McKinsey and Cognition pair Devin with consulting to prove AI coding ROI, as a survey shows productivity doubled at top firms but fell at nearly 30%.
By Olivia Hart
3 min read
What's News
- McKinsey and Cognition announced a partnership on Monday pairing McKinsey consulting with Cognition's AI coding agent Devin.
- A McKinsey survey of 300+ product and engineering leaders found the top quarter of organizations more than doubled AI-driven productivity, while nearly 30% saw it drop.
- One Cognition client compressed a five-year ERP modernization into 18 months, the company said.
- Goldman Sachs, Cisco, and JPMorgan have become more selective about AI tool access and usage.
McKinsey & Company and the AI lab Cognition have struck a partnership, announced Monday, built on a single proposition: that AI coding tools can pay for themselves — and that someone finally has to prove it.
Under the deal, McKinsey will advise clients on which engineering projects to target and how to redesign teams and workflows around AI. Cognition will supply its AI coding agent, Devin, which the company says can understand, plan, write, and test code. Neither firm disclosed financial terms of the arrangement.
Devin's launch in early 2024 rattled software engineers who feared AI tools would upend their workflows and perhaps overtake their livelihoods. In the years since, developers' work across industries has shifted dramatically — from writing code to guiding AI's output.
How much can AI actually deliver?
The honest answer, by McKinsey's own data, is: it depends on execution, and often it disappoints.
In a recent McKinsey survey of more than 300 product and engineering leaders, the top quarter of organizations saw productivity more than double due to AI. Yet nearly 30% of respondents said productivity had actually dropped. That spread — doubling for some, decline for nearly a third — is the real story behind the partnership.
Lari Hämäläinen, a senior partner in McKinsey's technology, media, and telecommunications practice, said the biggest changes for developers have come in the past year. Until the final months of 2025, many AI coding tools were most useful for narrow tasks, such as looking up technical guidance or syntax details. "That was about it," Hämäläinen said.
Since then, more capable AI models have made it possible to hand off longer, more autonomous coding work. But capability can outrun absorption. The technology might be improving faster than organizations can redesign the processes around it, Hämäläinen cautioned.
"This is really a step-change phenomenon" that "requires a complete rethink" of software- and product-development processes, Hämäläinen said.
The reinvention is not easy. While a small team can experiment with an AI tool, "it's a completely different thing to scale it across 2,000 developers or 20,000 developers," he said.
Why the Ferrari comparison matters
The hunt for AI returns — and ways to manage AI costs — has pushed large companies including Goldman Sachs, Cisco, and JPMorgan to become more selective about who gets access to which tools and how they are used. Cost discipline, not enthusiasm, is now the buying mode.
Art Levy, VP of global partnerships at Cognition, said Devin can draw on multiple AI models, using pricier frontier tools for harder engineering work and cheaper models for routine tasks.
"You don't need to drive a Ferrari to the grocery store," Levy said.
AI routing tools — which direct different tasks to models based on their capabilities and costs — are becoming a popular way for companies to manage AI spend. Devin's multi-model approach folds that logic into the agent itself.
Levy pointed to one client where work to update enterprise resource planning software, expected to take five years, is being compressed into 18 months. That is roughly a 70% cut in timeline, assuming the projection holds; the claim is Cognition's, and the client went unnamed.
"It's no longer, 'Can you get the job done?'" Levy said. "It's, 'Can you get the job done, and are you cheap at doing it?'"
What comes next
The partnership is, in effect, a bet that consulting guidance can close the gap between AI's potential and its messy rollout record. McKinsey's own survey says that gap is wide: for every organization doubling output, nearly one in three is going backward.
The unanswered question is whether process redesign — McKinsey's side of the deal — can be delivered at the speed the AI vendors promise. Hämäläinen himself flagged the risk that the technology is moving faster than the organizations adopting it.
Source: feeds.businessinsider.com
More from Olivia Hart
Show full bio
Staff writer covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.
2,205 articles