A survey of 525 leaders of organizations with more than 1,000 employees in the U.S. published this week finds nearly three quarters (72%) work for organizations that are already actively using or testing artificial intelligence (AI) agents, with 40% reporting they have deployed multiple agents in production environments.
Conducted by the market research firm Centiment on behalf of Zapier, a provider of an automation platform, the survey finds another 32% are still in the pilot or testing phase, while 14% say their organization plans to adopt AI agents even if they don’t have any right now.
Customer support (49%) and operations (47%) are the departments most likely to have deployed AI agents, followed by engineering (35%) and marketing (31%), sales (26%) and finance (24%).
Nearly half of respondents (47%) noting their organization is using AI agents to perform some type of data management task, such as data entry and extraction, followed by document analysis and summarization (41%), customer support triage and response (41%), and report generation (36%).
A full 84% of respondents also said it is either likely or certain their organization plans to increase investments in AI agents over the next 12 months, with 30% seeing automating routine workflows as having the most potential for AI agents. Only 2% say they’re certain they won’t increase investments in AI agents, while 1% said their organization is actively decreasing funding next year.
While the level of adoption and reliance on AI agents will naturally vary from one organization to the next, it’s clear there are employees that are developing the skills required to manage multiple AI agents, says Emily Mabie, an AI automation engineer at Zapier.
However, only 38% of respondents identified human-in-the-loop workflows as the most common approach to AI agent management. A total of 20% said their AI systems now operate autonomously with minimal oversight and only 18% identified security and data privacy concerns as the biggest barriers delaying AI agent adoption.
Nevertheless, organizations that don’t keep humans in the loop need to be wary of being dependent on a probabilistic set of AI tools that typically don’t perform the same task the same way twice, notes Mabie. While there are many tasks that can benefit from AI automation, more deterministic workflows that need to be completed the same way every time are going to lend themselves better to existing automation frameworks, she adds. “There are risks,” says Mabie.
Regardless of those risks, however, organizations appear to be aggressively moving forward. Just over half of respondents (53%) said they are using major cloud provider platforms such as Amazon Web Services (AWS), Microsoft and Google to build AI agents, while others are also relying on open-source tools (48%), enterprise AI platforms (48%), or orchestration frameworks (46%). Only just over a quarter (26%) said they are coding custom agents from scratch.
It’s obviously still early days so far as adoption of AI agents in the enterprise is concerned, but early returns make it clear that there is an opportunity to increase productivity that is simply too big to ignore. The only thing that now needs to be determined is how best to manage the inevitable disruption to existing processes and workflows that is about to inevitably ensue.
