AI Agents Need to Be Managed as If They Are Human

Agents Need Organization, Too

The Rise of Agentic AI

This year’s hot topic is agentic AI, and the specific tasks or services that AI agents can accomplish with minimum human intervention. The question is, how prepared are organizations to make AI agents work as intended?

The State of Agentic AI

A recent KMPG survey shows that more than half (51%) of organizations are exploring the use of AI agents, while another 37% are piloting AI agents. However, only 12% have deployed AI agents for use. Agentic AI may provide a path forward to returns on AI investments. "Faced with growing economic pressures, investor expectations, and the need for greater efficiency, business leaders are earnestly working to match their genAI goals with feasible implementation strategies," the KMPG authors stated.

Preparing for Agentic AI

Companies and their technology teams "are not quite ready, but there’s appetite from both technology teams and business teams to lean in," said Rodrigo Madanes, global innovation AI officer for EY. Two stages are necessary to prepare: formulating an agentic AI strategy and setting up infrastructure to implement agents – an agentic AI framework – following compliance processes.

Empowering, Delegating, and Verifying AI Agents

Michele Franceschini, head of AI communications engineering at Bloomberg, emphasized the importance of empowering, delegating, and verifying AI agents. "Empowering AI agents means ensuring they have access to the key elements they need to operate effectively," he explained. "Common examples are data and analytics. Are your data and analytics accessible programmatically? Is your data well-governed, clean, and managed scalably?"

Delegation and Verification

Delegation, another fundamental management tenet, is also essential to managing organizations of AI agents. "Delegation requires trust, and one key difference with agentic AI workflows is that AI agents are not perfect and are nondeterministic," said Franceschini. "The goal then becomes to build a robust delegation network utilizing imperfect, nondeterministic agents and a key challenge is that most engineering organizations require a cultural shift to accomplish this."

The Promise of Agentic AI

When properly managed, AI agents can prove to be really good workers. "Imagine a world where your best employee amplifies your work, learns faster than anyone else on your team, and helps you deliver assets like videos with unmatched speed and precision," said Chris Savage, CEO and co-founder at Wistia. "That’s the promise of agentic AI – digital teammates designed to learn from you, adapt to your needs, and revolutionize the way work gets done."

The Need for Cultural Shift

At the same time, a cultural shift is needed to help the productivity of AI agents measure up. "I believe most companies are not ready to support agentic AI because it fundamentally changes the workflow and how that set of actions is handled between humans and machines," said Arnab Mishra, CEO at Xactly. "I also think that most employees have not been thoroughly educated yet on how to truly understand how AI can help them do their jobs better, and thus are likely to be resistant to change."

Conclusion

Agentic AI has the potential to revolutionize the way work gets done, but it requires a cultural shift and a well-thought-out strategy. Organizations must empower, delegate, and verify AI agents to ensure they are working effectively. "The winners will be the companies that use it to go faster. The ability to delegate routine tasks to AI so humans can focus on work that requires creativity, strategy, and empathy will be a massive differentiator," said Savage.

FAQs

• What is agentic AI?
  Agentic AI refers to the use of artificial intelligence agents that can accomplish specific tasks or services with minimum human intervention.
• How prepared are organizations for agentic AI?
  According to a recent KMPG survey, more than half of organizations are exploring the use of AI agents, while another 37% are piloting AI agents. However, only 12% have deployed AI agents for use.
• What are the key elements to prepare for agentic AI?
  Empowering, delegating, and verifying AI agents are crucial to ensure they operate effectively.