Most enterprises now run AI agents in production, survey finds
Adoption jumped in a year. Autonomy did not.
Zapier's second annual survey of more than 500 enterprise leaders shows how quickly AI agents moved from trial to daily use. 71% said their organization has agents in production, up from 40% a year earlier. Another 17% are still testing.
Operations (59%) and customer support (57%) lead by department, followed by engineering, HR and finance. Marketing sits at 39%.
The more telling number is about autonomy. Only 30% of leaders whose companies use agents said those agents handle most of their tasks without a person stepping in. Most agents in production today do part of a job and hand the rest back.
The obstacles have not changed much. Data privacy, security and compliance topped the list at 27%, followed by accuracy and reliability at 24%. Even so, 46% said their most mature agent has paid for itself or will within 12 months, and 90% expect to spend more on agents next year.
Our read
The pattern makes sense. The first agent that sticks is usually narrow: sorting tickets, drafting a reply, pulling together a weekly report. The ones that stall tend to start with a broad mandate and no clear owner.
If you are planning a first deployment, pick one queue, one measure of success and one person who reviews the output every week.