2 days ago

McKinsey Finds Agentic AI Productivity Declines At 30% Companies

McKinsey Finds Agentic AI Productivity Declines At 30% Companies
McKinsey report shows productivity decline in 30% of companies using agentic AI · livemint.com

A McKinsey report studied how companies use agentic AI tools for software development.

These tools can perform coding tasks with limited human direction.

However, nearly 30% of companies saw productivity go down after adopting them.

In one study, the amount of coding increased a lot, but the number of completed product releases increased much less.

This shows that writing more code does not always create more useful products.

Many developers also worry that AI tools are not accurate.

Investment in companies developing these tools has grown very quickly.

The report says companies need organized processes and developer trust to gain value from agentic AI.

Key facts

Productivity decline
Nearly 30% of companies saw productivity fall after adopting agentic AI tools.
Coding activity
One study recorded a 180% increase in coding activity.
Shipped releases
Shipped releases increased by only 30% in that study.
Developer trust
Forty-six percent of developers distrusted AI accuracy, 33% trusted it, and 3% highly trusted outputs.
Equity investment
Investment grew from negligible levels through 2023 to about $5 billion in 2025 and more than $61 billion in the first half of 2026.
Job postings
Relevant job postings rose 221% between 2024 and 2025.
Other technology investment
AI infrastructure spending doubled in one year, while energy technologies attracted nearly $200 billion during 2025.

Quotes

McKinsey’s Technology Trends Outlook 2026

McKinsey report on technology trends and enterprise AI adoption

“In one study, AI tools increased coding activity by 180 per cent, but shipped releases rose by only 30 per cent, showing that more code doesn't always lead to more products.”
livemint.com
“Some 46 per cent of developers worldwide actively distrust AI tools' accuracy, compared with 33 per cent who trust them, and only 3 per cent highly trust their outputs.”
livemint.com

Sources

Related news