Software review site G2 released its 2026 AI Agents Report on Oct. 14. It found that 66% of organizations now have AI agents in limited or full production, and that the share of business software buyers who bought agent software or services rose from 69% to 86% in a year.
The report, titled “Crossing the Agentic Chasm,” is based on a survey of more than 1,000 business software buyers plus G2 review data, according to G2’s announcement. The findings come from G2’s own research.
Tim Sanders, G2’s chief innovation officer, wrote that the largest group of buyers, including business process owners and IT, is “wired for proof, not promises.”
The oversight cost
G2 calls the problem a “distrust tax,” which it defines as the ongoing cost of running agents that companies don’t fully trust. It found that 31% of business buyers say employees spend six or more hours a week “bot-sitting,” its term for overseeing agent work.
And 45% say time spent overseeing and correcting agent mistakes eats up more than a quarter of the productivity gains they expected, according to the report. Another 46% report higher-than-expected operating costs.
The burden falls harder on big companies. G2’s press release says 27% of larger enterprises require significantly more oversight, nearly twice the 15% rate at small and midsize businesses.
“AI agents are showing real signs of value, but much of it is being erased by oversight, rework, and exception handling,” Sanders said in the release.
There are gains to measure. Four in five organizations using agents in high-volume workflows report cycle-time improvements of at least 11%, G2 says. Some 83% say agents improve employee satisfaction by cutting repetitive work, and just 14% report reducing headcount.
Too many agents to track
G2 says agent fleets are growing faster than companies can manage them. Of organizations with agents in production, 38% run more than 100 of them, yet 47% still manage them informally or through individual departments.
Just 27% keep a central registry of their agents, and 14% have real-time monitoring of unusual behavior, costs or risks, according to the release.
The survey also points to pricing pressure on software vendors. Only 9% of buyers say per-seat pricing is fairest for agents, Sanders wrote, and 72% call a strong set of connectors and add-ons very important or a must-have.
Buyers are cautious about checking agents’ work. G2 found 52% require human quality review before an agent goes into production, and 77% re-evaluate agent performance at least monthly.
A separate G2 analysis of 3,222 reviews from enterprise buyers found a similar pattern. It said 77% of enterprise reviewers trust their AI agent, but only 7% would give it total autonomy.
Looking ahead, 84% of buyers say agent capabilities will influence their 2027 software choices, the release says.
Source: G2, Tim Sanders, “New G2 Research: AI Agents Have Moved From Pilot to Production. Now Comes the Distrust Tax.”
Additional sources: PR Newswire, G2, “New G2 Research: AI Agents Have Moved From Pilot to Production, But Oversight Comes at a Cost”; G2 Learn, G2, “Stop Selling Autonomy: What 3,222 Enterprise Reviews Say AI Agent Buyers Want”
