AI adoption is accelerating faster than organisational control

What the 2026 evidence shows about how widely AI is used, and how slowly the right to act is being governed.

10 September 2026 · 7 min read

There is a gap between using a tool and granting it authority. Public research in 2026 is now large enough to show that gap in numbers, without treating any publisher as a partner of Solarascope.

What the sources report

Source finding

Stanford HAI's 2026 AI Index (Economy) reports organisational AI adoption at 88% of surveyed organisations in 2025, with generative AI used in at least one business function at 70%. Agent deployment remained in the single digits across nearly all business functions.

McKinsey's State of AI in 2026 finds nearly nine in ten respondents using AI regularly in at least one function. Forty percent of respondents from large organisations (over US$1 billion revenue) report scaling AI agents, up from 27% a year earlier. Scaling in smaller organisations stayed flat at 22%.

McKinsey's 2026 AI Trust Maturity Survey reports average responsible-AI maturity moving from 2.0 to 2.3, while strategy, governance and agentic controls lag, with only about 30% of organisations at maturity level three or higher on those dimensions.

Solarascope analysis

Adoption is not control. An organisation can have a chatbot in every function and still have no inventory of which systems may spend money, change records or commit the firm. The Stanford single-digit agent deployment figure and McKinsey's lagging agentic controls are the same story told from two sides: the right to act is spreading faster than the record of why it was allowed.

That is the job of reliance infrastructure. Not another policy. A current, named, bounded basis for each consequential grant of authority.

Sources

Stanford HAI. AI Index Report 2026: Economy (April 2026)
McKinsey. The state of AI in 2026: On the road to ROI (August 2026)
McKinsey. State of AI trust in 2026: Shifting to the agentic era (March 2026)