Insights ·
Who runs AI agents in 2026, and where a small business can start
Who runs AI agents so far per McKinsey and BCG, Census data on AI use by firm size, and why a small business might start at Gartner's advise level.
McKinsey’s survey defines AI agents as systems that act in the real world and can plan and carry out several steps of a workflow on their own. For a small business that might be software that answers quote requests from your inbox with your prices and sends them out without anyone on your team looking first.
Who uses agents now
McKinsey’s 2026 survey had 1,719 respondents in 97 countries. At organizations with more than $1 billion in revenue, 40% report scaling AI agents, up from 27% a year earlier. At smaller organizations it stayed at 22%. Smaller in that survey means under $1 billion in revenue, which still includes companies far bigger than most owner-run businesses.
The Census Bureau does cover small firms, but it asks about any use of AI and not agents. In the collection period ending 3 May 2026, 32% of US firms with 100 to 249 employees had used AI in a business function in the past two weeks. The national rate was 19.8%. From December 2025 to May 2026 AI use rose at firms with at least 20 employees, and at firms with fewer than 20 there was no significant change.
BCG surveyed workers, and 30% of them say AI agents are integrated into workflows, up from 13% the year before.
Gartner’s two forecasts
In June 2025 Gartner predicted that over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls. In May 2026 it said that by 2027, 40% of enterprises will demote or decommission autonomous AI agents because of governance gaps found only after production incidents. Half of BCG’s respondents say their company lacks clear governance for teams of people and AI.
Starting with AI agents in a small business
Gartner also sorts agents into four levels of autonomy: observe, advise, act with approval, and act autonomously. At advise, people review everything the agent produces and carry out the actions themselves. We’d start there, because at that level a wrong quote can be caught by the person who reads it before it goes to the customer. The agent drafts, and someone on your team reads it, edits it if they need to and sends it. If you want to talk through which task to try first, you can book a free 30-minute fit check with us.
Before it goes live, pick the person who answers for what the agent does, and make sure they and one other person know how to switch it off. After that it’s worth counting how many drafts go out without changes. If that stays high for a few weeks you could move up to act with approval, and keep a record of each approval so you can see later who signed off on what.
Sources
- McKinsey, The state of AI in 2026: On the road to ROI, 25 August 2026.
- US Census Bureau, Large Firms With at Least 20 Employees Biggest AI Users, 26 May 2026.
- BCG, AI Is Reshaping Jobs Faster Than Companies Are Reshaping Work (press release), 3 June 2026.
- Gartner, Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, 25 June 2025.
- Gartner, Gartner Says Applying Uniform Governance Across AI Agents Will Lead to Enterprise AI Agent Failure, 26 May 2026.