
PRODUCTIVITY DECLINE
~30%
TOP QUARTILE GAIN
2×+
LEADERS SURVEYED
300+
ERP COMPRESSION
5yr→18mo
McKinsey and Cognition, the company behind AI coding agent Devin, announced a partnership aimed at proving that AI-assisted software development actually delivers returns. McKinsey will advise clients on which engineering projects to target and how to restructure teams around AI tools. Cognition provides Devin, which the company says can understand, plan, write, and test code autonomously.
Extraction Map
The buried headline is in McKinsey's own data. A survey of over 300 product and engineering leaders found that the top quartile of organizations more than doubled productivity through AI — but nearly 30% of respondents reported productivity actually declined. That is not a rounding error. It means the median outcome of AI coding adoption is far less impressive than the highlight reel suggests. Lari Hämäläinen, senior partner in McKinsey's tech, media, and telecom practice, frames the challenge as a pace mismatch: AI tool capabilities are advancing faster than organizations can redesign their surrounding processes. Until late 2025, he says, most AI code tools were limited to narrow tasks like querying technical documentation or syntax details. Since then, more capable models have enabled longer, more autonomous coding work — but that leap creates organizational complexity that small pilot teams never encountered. The scaling problem is real. Hämäläinen notes that piloting AI tools with a small team is easy, but extending them to 2,000 or 20,000 developers is "a completely different matter." This is where McKinsey's consulting revenue enters the picture. The partnership is structured so that McKinsey sells the strategy and change management, while Cognition sells the tool — a classic bundling play where the consulting engagement creates lock-in for the AI product and vice versa. Cost management is becoming central to the AI coding conversation. Goldman Sachs, Cisco, and JPMorgan are already restricting who can use which AI tools and how. Cognition's Art Levy acknowledges this directly: "You don't need to drive a Ferrari to the grocery store." Devin can route tasks to different AI models based on difficulty and cost — an approach that mirrors the broader rise of AI routing tools across enterprise. One client case study stands out: a company originally projected to spend five years modernizing its enterprise resource planning software is now compressing that timeline to 18 months using Devin. If real, that is a significant acceleration. But ERP modernization projects are notorious for cost overruns and scope creep regardless of tooling, so the 18-month claim deserves scrutiny over time. The deeper dynamic here is that McKinsey is positioning itself as the indispensable middleman in AI adoption — the firm that tells you which AI to buy, how to reorganize around it, and measures whether it worked. That is a consulting model as old as consulting itself, now applied to the hottest budget line in enterprise tech. The question is whether the value created flows primarily to the enterprises adopting AI, or to the consultancy-vendor axis selling the transformation.
Worth reading?
OPTIONAL
Score
40
/100
Extractive leaning — the partnership bundles consulting fees with AI tooling in a structure where McKinsey captures value whether client productivity rises or falls.
What Is GRIN?
GRIN stands for Generativity, Resilience, Innovation efficiency, and Novelty openness. These four dimensions capture the essential characteristics of any memetic intelligence — whether a nation, a corporation, a religion, or a movement.
About GRINGenerativity
2/5
Resilience
2/5
Efficiency
2/5
Novelty
3/5
Extractive
4/5
Inverted
Does this create new capability or just redistribute existing value?
Potential is real but unproven at scale. McKinsey's own data shows most adopters see flat or declining productivity. The generative case rests on a top-quartile minority.
Verified Facts (8)
McKinsey survey of 300+ product and engineering leaders found top 25% of organizations more than doubled productivity through AI
Nearly 30% of respondents reported productivity declined after AI adoption
Lari Hämäläinen is senior partner in McKinsey's tech, media, and telecom practice
Art Levy is Cognition's VP of global partnerships
One Devin client compressed a 5-year ERP modernization to 18 months
Devin launched in early 2024
Goldman Sachs, Cisco, and JPMorgan are being more selective about AI tool access and usage
Hämäläinen stated that until the last months of 2025, most AI code tools were limited to narrow tasks
More latest news
Read all