Knowledge management is the second most common place in the enterprise where AI agents have reached the scaling phase. Only IT ranks higher, and the gap between them is a single percentage point.
The figures come from a McKinsey report The state of AI in 2026: On the road to ROI.
Across all respondents, 12 per cent report scaling agents in IT and 11 per cent in knowledge management. Software engineering follows at 10 per cent. Manufacturing sits last at 3 per cent.
The online survey ran from 4 May to 8 June 2026. It drew 1,719 responses from 97 nations, spanning industries, company sizes, functions and seniority levels. Thirty-six per cent of respondents work for organisations with more than US$1 billion in annual revenue. Results are weighted by each respondent nation's contribution to global GDP.
Agents are defined in the report as "AI systems based on foundation models that act in the real world and are capable of autonomously planning and executing multiple steps in a workflow". The function-level questions were asked only of respondents reporting regular AI use in that function, then rebased to the full sample.
After IT, knowledge management and software engineering come service operations and product or service development, both at 9 per cent. Marketing and sales sits at 8 per cent and human resources at 7 per cent.
Three functions share the bottom of the table at 5 per cent. They are risk, legal and compliance; strategy and corporate finance; and supply chain or inventory management.
The report reads the pattern as organisations scaling agents in the functions carrying the greatest potential value for their industry. The industry breakdown supports that. Almost every sector shows one or two functions well above its own baseline.
Coding agents change what gets bought
The agent numbers connect to a procurement shift. Nearly a third of organisations, 32 per cent, have decided against buying at least one software product or feature. They concluded they could build it in-house with agentic coding tools instead.
The report describes that as a possible sign that AI is starting to reshape how technology budgets are allocated. About two in ten organisations are already scaling software coding agents, rising to 31 per cent at larger enterprises.
Nearly nine in ten respondents report regular use of AI in at least one business function. The share reporting AI scaling across the enterprise rose from 38 per cent to 44 per cent in a year. The share using AI in three or more functions rose from 51 per cent to 56 per cent.
Among organisations with at least US$1 billion in annual revenue, 54 per cent report scaling AI across the enterprise. Among smaller organisations the figure is a third.
The gap is wider for agents. The share of large organisations scaling agents in one or more functions rose from 27 per cent to 40 per cent over the year. Among smaller organisations it stayed flat at 22 per cent.
About 20 per cent of respondents said AI-related operating costs, including token costs, constrained their organisation's AI use. Most still expect their organisations to increase AI investment over the year ahead.
Eighty per cent of respondents report that AI has improved their individual productivity. Half say it helps them make better decisions.
Enterprise financial impact has not followed. The share attributing at least some EBIT impact to AI use sits at 37 per cent, essentially unchanged from a year ago.
Thirty-nine per cent of respondents expect AI-related declines in their organisation's total employment over the coming year, up from 32 per cent last year. Another 43 per cent expect little or no change.
Last year's expectations proved well ahead of events. Among respondents from organisations using AI, just 14 per cent said AI contributed to an overall decline in workforce size last year. That is less than half the 32 per cent who expected reductions a year ago.
McKinsey associate partner Tara Balakrishnan writes in the report that the limiting factor is increasingly the organisation's ability to absorb change. The challenge, she says, is "to avoid simply accumulating tools or pilots and instead to build a repeatable capability to transform these efforts into new ways of working".