Clearing a Path to Success With Agentic AI in Cybersecurity

PwC's Alex Cherones and Marianne Olsen on Managing Risks, Improving Cyber Defenses Cal Harrison • June 16, 2025

In the early days, up to 98% of artificial intelligence-related projects failed. Today, with the emergence of agentic AI with greater autonomy and ease of use, organizations are finding much higher success rates - if they clearly define business use cases and mitigate the associated risks, said PwC's Alex Cherones and Marianne Olsen.

See Also: How Agentic AI Presents New Identity Security Challenges

Many of those ill-fated AI projects emerged out the buzz around generative AI - and now agentic AI - but "companies were just choosing tons of different use cases, and weren't necessarily thinking it through," said Olsen, director of cybersecurity and risk management at PwC. Rather than aiming for a moon shot, organizations should look at smaller, tightly defined projects and build on those successes, she advised.

"We're seeing ...