A survey of 1,000 business and IT executives published today finds more than half (51%) lead organizations have deployed artificial intelligence (AI) agents, with another 13% planning to deploy them within the next year.
Conducted by Wakefield Research on behalf of PagerDuty, the survey also finds that 52% of respondents are also leveraging agentic AI to automate or accelerate 26% to 50% of their workloads.
Additionally, 62% of respondents anticipate a return on investment (ROI) on agentic AI of 100% or more, with 75% of organizations planning to invest in excess of $1 million in AI initiatives.
The pace of adoption of AI agents shows that initial investments in generative AI have laid a foundation that is making it simpler for organizations to embrace agentic AI, says PagerDuty CIO Eric Johnson.
In fact, the survey finds a full 94% of respondents expect adoption of agentic AI will be faster than initial generative AI technologies.
The challenge now is determining which AI agent projects to prioritize given limited bandwidth and IT budgets, notes Johnson. The issue is that as business executives are exposed more to AI agents, the number of potential use cases continues to exponentially increase, he adds.
Hard generative AI lessons already learned that organizations are hoping to not repeat include: Rushing to deploy (41%), spending too much (40%), not setting up guidelines (38%), not offering sufficient training (37%), not having the right data infrastructure (37%), not setting the right expectations (36%) and spending too little (35%), the survey finds.
In general, the one thing that most agentic AI initiatives have in common is they generally reduce dependencies on humans to provide Level One support services, says Johnson. Most of the individuals performing those tasks would prefer to focus more on higher value tasks. The challenge and opportunity that organizations now face is deciding whether, as those tasks are automated, to either reallocate that human capital to other functions that drive additional revenue or simply drop that savings to the bottom line, he adds. “We’re going to have to rethink human capital,” says Johnson.
Many organizations may also be underestimating the level of change management that agentic AI will require as tasks spanning multiple functions are automated, notes Johnson. AI agents, for example, will be able to automate a range of sales and marketing tasks that were previously managed by separate teams, he adds.
The fact is that when it comes to AI agents, there are more unknowns than knowns. As large language models (LLMs) continue to expand their reasoning capabilities, the types of tasks that AI agents will be able to complete will become more complex. At the same time, orchestration frameworks will make it possible to use multiple AI agents to automate tasks on an end-to-end basis.
Humans will always be needed to ensure those processes are being reliably automated, but the one thing that is certain is that, for better or worse, much of the drudgery that makes work unpleasant to perform today is about to be eliminated.
