Enterprise

McKinsey and Cognition partner to prove AI coding ROI

The consulting giant will help enterprises redesign workflows around Devin, Cognition's autonomous coding agent, as companies struggle to scale AI tools profitably.

Omega Editorial · November 16, 2026 · 3 min read

Key takeaways

  • McKinsey survey found top-performing organizations doubled productivity with AI coding tools, while 30% saw productivity decline.
  • The main adoption challenge is scaling AI coding from small experiments to thousands of developers, requiring complete workflow redesign.
  • Cognition's Devin routes tasks across multiple AI models to balance capability and cost, using expensive models only for complex work.
  • One client compressed a five-year ERP modernization project to 18 months using AI coding agents.
  • Major enterprises including Goldman Sachs and JPMorgan now restrict AI tool access to manage costs and prove ROI.

McKinsey tackles the AI coding adoption gap

McKinsey & Company has formed a partnership with AI lab Cognition to address a critical challenge facing enterprises: proving that AI coding tools deliver measurable returns on investment. Under the arrangement announced Monday, McKinsey will advise clients on selecting appropriate engineering projects and restructuring teams and workflows to accommodate AI agents, while Cognition provides Devin, its autonomous coding tool that can understand requirements, plan solutions, write code, and run tests.

The collaboration comes as organizations grapple with a widening gap between AI capabilities and their ability to deploy those tools effectively across large engineering teams. According to Lari Hämäläinen, a senior partner in McKinsey's technology, media, and telecommunications practice, AI coding tools remained limited to narrow tasks like syntax lookups through most of 2024 and into 2025. More capable models emerging in late 2025 changed that equation, enabling longer autonomous coding sessions that require fundamental rethinking of software development processes.

Why it matters

The partnership reflects a broader enterprise challenge: AI coding tools are advancing faster than companies can adapt their workflows. McKinsey's recent survey of over 300 product and engineering leaders revealed stark performance divides—the top quartile of organizations more than doubled productivity with AI, while nearly 30 percent reported productivity declines. This disparity suggests that technology alone won't deliver returns; successful adoption requires systematic organizational redesign that many companies lack expertise to execute.

Scaling remains the sticking point

While small-team experiments with AI coding tools prove straightforward, expanding those pilots to thousands of developers presents distinct obstacles, Hämäläinen noted. The challenge extends beyond technical integration to encompass team structures, approval processes, code review practices, and quality assurance workflows.

Cost management has emerged as another friction point. Major financial institutions and technology companies including Goldman Sachs, Cisco, and JPMorgan have implemented selective access policies, controlling which employees can use which AI tools for specific tasks. Art Levy, vice president of global partnerships at Cognition, said Devin addresses cost concerns by routing tasks across multiple AI models based on complexity—reserving expensive frontier models for difficult engineering problems while using cheaper alternatives for routine work.

Compressing multi-year projects

Levy pointed to practical applications where the economics prove compelling. One client undertaking enterprise resource planning software modernization originally projected a five-year timeline. Using Devin, the company expects to complete the work in 18 months. Such dramatic compression of long-deferred technical debt projects represents the kind of ROI case that can justify AI coding investments.

The shift in developer expectations reflects this new reality. "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?'"

Details of the partnership were first reported by Business Insider.

  • #ai coding
  • #cognition
  • #mckinsey
  • #devin
  • #developer productivity
  • #enterprise ai

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Enterprise