A survey of 1,000 IT decision-makers in organizations with more than 250 employees in the United States, United Kingdom, Germany and Australia finds 79% work for organizations that plan to make agentic artificial intelligence (AI) a high priority by investing more than $1 million over the next 12 months.
Conducted by 3GEM Research on behalf of SnapLogic, the survey also finds 92% of respondents are either very confident (50%) or somewhat confident (42%) that AI agents will drive real business results within 12-18 months, with AI agents expected to reduce the time to complete a task on average by 19 hours per week.
In fact, a full 84% said they trust AI agents as much (44%) or more (40%) than humans to complete an assigned task.
Overall, the survey finds 90% of respondents report their organization is already using generative AI technologies, with 50% claiming to be using AI agents. Another 24% plan to in the next 12 months, while 18% noted their organization has in the past but stopped.
Major barriers to adoption cited by all respondents include data security and privacy concerns (60%), legacy technologies and lack of integration (46%), lack of employee understanding (29%) and fear of AI hallucinations (14%).
Despite these concerns, most respondents are moving forward with agentic AI investments, but there is a significant difference between a greenfield application and one that seeks to apply agentic AI to an existing business process, says SnapLogic CTO Jeremiah Stone. Most existing business processes over time evolve to incorporate a number of exceptions to the rules that were initially applied via one or more applications. Many organizations, as they move to adopt agentic AI, will find they will first need to reengineer those workflows in a way that lends itself to be completed by an AI agent, notes Stone. “There will be a rebirth of business process re-engineering,” he says.
It’s unclear to what degree organizations will rely on consultants as they have in the past to drive those initiatives but it’s already apparent that some level of orchestration will be needed to complete an end-to-end business process that will be performed by AI agents working alongside humans, adds Stone.
The challenge is that many of the processes that organizations are trying to automate are deterministic in the sense that they need to be completed the same way every time. Generative AI tools today are probabilistic, in that, while providing higher levels of accuracy, still don’t complete tasks the same way every time. However, in time as AI agents are trained with a narrow set of data, they will complete tasks in a more deterministic manner, Stone notes.
At this juncture, it’s not so much whether AI agents will be incorporated into workflows so much as it is how, when and to what degree. The simple truth is many of these agents will be performing tasks that many humans really enjoy doing in the first place. The real challenge then becomes determining how best to apply the ingenuity that humans uniquely possess to add more value above and beyond what an AI agent might be able to do on its own.
