Jonathan Kitchen | Digitalvision | Getty Images
Companies are throwing lots of money into artificial intelligence tools and services, particularly agentic AI, in many cases without truly understanding what these offerings actually do. A key task for technology leaders going forward is to alter this scenario before more investments are wasted.
“It’s not surprising that many organizations are still trying to figure out what agentic AI really is,” said Dan Diasio, global AI leader at consulting firm EY. “Agentic AI has a buzz about it that many in the market want to capitalize on, and we’ve seen an incredible rebranding of anything related to generative AI presented as ‘agentic AI.'”
A lot of today’s gen AI use cases are what Diasio would call assistants, where a user types a prompt and the tool provides a response, suggests recommended actions, or handles administrative tasks.
“But an agent has greater autonomy, recognizing when a task should be done and completing all the steps themselves while understanding the context of the situation,” Diasio said. “Both are useful, but the latter is vastly more powerful and aligned with the true potential of AI.”
As part of its latest U.S. AI Pulse Survey, EY queried decision-makers across a range of industries, and found that 21% of senior leaders say their organizations have invested $10 million or more in AI, up from 16% a year ago. About one-third anticipate spending $10 million or more on AI next year.
There is a significant divide between senior leaders’ commitment to agentic AI and its full-scale adoption, however. While early agentic AI implementations are delivering tangible benefits, only 14% of respondents said agentic AI technology has been fully implemented in their organization.
“The survey shows that most organizations are not yet prepared for agentic AI’s demands,” Diasio said. “This includes having organized, high-quality knowledge to guide these systems and a clear understanding of how companies navigate the massive change between the current and future states.”
‘Climate of uncertainty’
As a result of this lack of preparation, many companies have promising opportunities but are hesitant to move beyond pilot programs, Diasio said. That’s the case even with successful returns from current AI investments.
“While this combination of technical unpreparedness and change readiness creates a climate of uncertainty, it provides a clear roadmap for organizations,” Diasio said. “By addressing these foundational issues first, companies can confidently move beyond pilot programs and bridge the gap between strategic commitment and full-scale, enterprise-wide implementation.”
The accelerated evolution of AI presents a significant challenge for enterprise adoption, said Deepankar Mathur, associate director at research firm Searce. “It’s an environment where many organizations, particularly those for whom technology is an enabler rather than a core product, feel they are perpetually behind the curve or reacting to the latest market development.”
Mathur says the concept of “full-scale adoption” is losing relevance in the context of agentic AI’s rapid evolution. “It’s like trying to hit a constantly moving target,” he said. “Enterprises need to shift their focus away from a singular, comprehensive implementation event.”
Instead, companies need a continuous optimization cycle that involves systematically identifying processes for automation, prioritizing them based on an impact-versus-effort analysis, and deploying the best available solution at that moment. “The key is that immediately after deployment, the process of refinement and enhancement begins again,” Mathur said. “For the foreseeable future, this cycle of improvement isn’t a temporary project; it’s an ‘always-on’ operational imperative.”
Human knowledge and AI assets
There are steps technology and business leaders can take to help their organizations thrive with AI.
One is to clearly define the human-AI partnership. Leaders should approach agentic AI integration as “a symbiotic relationship with existing talent,” Diasio said. “This means crafting a strategy that outlines what tasks AI will handle and what roles humans will play. [This] approach is a more effective and engaging way to leverage agentic AI, helping to alleviate employee fears and foster a more collaborative environment.”
Another good practice is to turn “tacit knowledge” into knowledge assets, Diasio said. “Jobs are performed through know-how and experience, which is information that may exist only in workers’ heads, not aggregated in historical databases,” he said. “Agentic AI needs this organizational knowledge to guide effective decision-making with a consistent methodology.”
With autonomous agents, the potential for both positive and negative outcomes is greatly improved and technology leaders need to address this.
“Recently, we’re starting to get more news of the cyber [security] implications in many agents, and that will only grow with more agents being pushed into production,” Diasio said. “Therefore, it is crucial to establish a responsible AI framework and a cyber plan optimized for AI from the outset.”
This involves setting clear policies on data privacy and security, ethical use, and oversight to determine the specific points where human review is mandatory, Diasio said. “By proactively addressing these governance questions, leaders can build a trustworthy and transparent system that aligns with company values, manages risk, and builds confidence among employees and stakeholders,” he said.
Even with the potential risks, businesses also need to democratize access to AI tools. “The barrier to AI implementation has significantly lowered; one no longer needs to be an [machine learning] engineer to create value,” Mathur said. “Forward-thinking organizations should prioritize democratizing access to agentic AI tools within a well-defined security framework.”
This empowers employees to be innovative within their daily tasks. “Over-reliance on centralized AI councils or steering committees can create bottlenecks and stifle this bottom-up innovation,” Mathur said. “True progress requires leaders to actively champion and enable this widespread adoption.”
Another good practice is to create an AI center of excellence. “The most successful enterprises are building small, elite teams of ‘AI blackbelt’ specialists who operate as a horizontal center of excellence,” Mathur said. “These experts embed directly within various business functions, not to do the work for them, but to enable and train those teams to build their own agentic workflows.”
In addition, businesses need to be smart about setting goals for AI. “Be specific with internal teams on desired outcomes, find ways to measure success,” Mathur said. “Objectives should be achievable, realistic and targeted for achieving within mutually agreed upon timelines by the team.”
