AI agents: High effort, low return
37% of companies see positive EBIT from AI. McKinsey's latest survey reveals why agent deployments are draining budgets faster than they're delivering returns.
Why it matters
McKinsey's 2026 Technology Trends Outlook documents a widening gap between AI adoption (89% of enterprises) and economic impact (37% report positive EBIT). Agent deployments consume 5–30× more compute than chatbots, causing 93% of companies to exceed AI budgets. Only 25% of software teams achieved significant productivity acceleration; 30% saw productivity decline. The finding reframes the agent opportunity: maturity and better tooling (LangGraph, improved Claude/OpenAI agents) are closing the gap, but current deployments show quality, trust, and workflow redesign matter more than code volume.
The key facts
11 to know89% of enterprises use AI regularly; only 37% report positive EBIT impact
Agent workflows require 5–30× more compute than typical chatbot queries
93% of surveyed companies exceeded AI budgets due to agent resource demands
Only 25% of companies using AI agents for software development achieved significant acceleration (≥2× productivity for >25% of teams)
80% of developers saw ~3% productivity gains; top 20% saw 55% gains
30% of companies saw productivity decline after deploying coding agents
Programming activity increased 180% with AI tools; published releases increased only 30%
46% of developers actively distrust AI tool accuracy; 3% have great confidence
Investments in agent-based software development solutions: ~$5B in 2025; >$61B in H1 2026 (driven partly by SpaceX's $60B Cursor acquisition)
Job postings for agent-related roles increased 221% between 2024 and 2025
METR benchmark: Claude Opus 4.6 estimated at 12-hour task horizon for software engineering (refers to human task duration, not AI processing time)
Go to the source
CIOcio.com
Publisher excerpt: AI has arrived in the business world — and the results are thus far underwhelming. According to McKinsey, 89% of companies now use AI regularly, and the majority are at least experimenting with AI agents. However, a significant gap remains between usage and economic success. “We find that only 37%…