Tech execs are getting wise about ROI from AI
80% say AI boosted productivity. Only 6% saw it hit the bottom line. Here's why the gap is widening — and what it costs to close it.
Why it matters
Enterprise AI ROI remains elusive despite widespread adoption; the gap between individual productivity gains and organizational profit impact is forcing CFOs and CIOs to redesign workflows and tighten cost controls, reshaping how companies evaluate and deploy agents.
The key facts
13 to knowMcKinsey 'State of AI' survey (August 2026): 80% report AI improved productivity, but only 37% saw profit impact, only 6% achieved significant value (≥5% of operating profit)
Gartner survey (Data & Analytics Summit, Mumbai): 60% of IT leaders worried about unexpected AI agent costs; some found coding-assistance token costs exceed human developer salaries
Gartner 2025 baseline: one in five AI initiatives achieved ROI
KPMG global survey (released last month): 55% of organizations have formal AI harness layer; 86% of those reporting established ROI do
Agent-to-agent cost risk growing; agents layered onto existing workflows insufficient without end-to-end redesign
Gartner introduced 'return on intelligence' framework: financial + non-financial outcomes
McKinsey finding: high-performing companies redesign entire workflows, not add agents to existing processes
80% of respondents report AI improved productivity; only 37% see profit impact; only 6% credit AI with 5%+ of operating profit (McKinsey State of AI survey, August 2026)
3 out of 5 IT leaders worry about unexpected AI agent costs; some organizations already find coding-assistance token costs exceed human developer salaries (Gartner Data & Analytics Summit, Mumbai)
55% of organizations have formal AI harness layer for governance; rises to 86% among those with established ROI (KPMG global survey, recent)
Only 1-in-5 AI initiatives achieved ROI in 2025 (Gartner)
McKinsey finding: high-performing companies redesign workflows end-to-end with agents, not just layer agents onto existing processes
Key cost control challenge: lack of transparency on which workloads drive token expenses; token costs for coding assistance in some cases exceed human developer costs
Go to the source
Computerworldcomputerworld.com
Publisher excerpt: Most organizations have been largely unable to measure financial returns from AI, but analysts say new ways to calculate return on investment are emerging. “There’s a delay between the development of technology, even the investment in the technology, and the value that an organization can capture…