How to implement agentic AI correctly as C-suite presses for it

IT professionals might be confused about the best ways to implement agentic AI. But with C-suite executives pressing for their organizations to adopt AI as quickly as possible, they need to make some kind of decision about AI agents. What’s their best course forward?

Agentic AI, while offering the tantalizing possibility of AI agents taking care of complicated workflows and saving everyone time, shouldn’t necessarily be applied to every single system or process without a deep evaluation.

That can lead to issues with managing C-suite expectations—your CEO wants agentic AI implemented now because they heard all about it at a conference.

John Searby, the chief strategy officer at HUMAN Security, told IT Brew that success stories of automation are making executives react.

“Everyone wants to look at the bottom line and say…I can find some sort of improvement there, if I can automate better,” Searby said. “That has been pushed out as mandates in a lot of places. And the question has been, what is the risk of adopting it, cost of adopting it, and do we actually need it?”

Experts like Krishna Sai, CTO at SolarWinds, said that looking at things like key indicators and internally articulating that IT teams are dealing with complexity across workloads are important places to begin.

Managing expectations. Searby agreed with Sai and said that the most effective way of communicating with the C-suite about agentic AI is starting with a key question: Is change really possible?

“Let’s start with an open mind—new technology can be scary, especially one that’s going to turn into a lot of work for you or maybe disrupt your job in some way,” Searby said.

Once a team learns more about agentic AI, testing and using its components is key, along with a proper risk assessment.

“Always to any executive mind, when you represent the risk side of it…there’s a cost associated with it, and you think that there’s cost efficiencies to be gained, but if I’m measuring this process and what if it goes wrong, imagine the potential loss on the other side,” Searby said.

Can we get another opinion? Cristian Rodriguez, the field CTO for the Americas at CrowdStrike, said the decision to implement agentic AI is beyond an IT professional’s control. He advised those in IT to advocate for ensuring agentic AI systems meet certain benchmarks for trust, governance, and resilience.

If the C-suite is really excited for agentic, Rodriguez suggests every business unit that’s using AI perform an impact analysis and understand how models interact with data.

Additional strategies could include bringing in a third-party consultant to provide expert-led, security-focused guidance on incorporating AI agents into various business units. For example, such a consultant could help the organization find the right balance of AI agents to humans, security and data requirements, and implementation timelines for agentic AI.

More importantly, Rodriguez said, an outside party can help these leaders convey the message of where the risk is and how the business can still be successful by adopting AI.

Before you implement agentic AI due to C-suite pressure, it’s also valuable to ask what they’re adopting the technology for, according to Searby. Teams should also evaluate if their existing IT infrastructure is capable of incorporating AI agents; upgrades could mean anything from changing existing cloud-service subscriptions to adding more storage.

“Not everything needs autonomy, and some things are riskier to inject that into,” Searby said. “That’s the question, do I need the autonomy? Will I see a marked improvement?”