CIOs Build Control Systems as Autonomous AI Agents Proliferate

Chief information officers are confronting a new operational challenge: autonomous AI agents that can take initiative, explore system vulnerabilities, and operate beyond direct human oversight. As these digital workers spread across enterprise functions from supply chain to customer service, technology executives are building centralized control systems to track and govern their behavior.

"This is a risk that enterprises need to be focused on, understand, and start planning for," Joe Atkinson, global chief AI officer at PwC, told Fortune, which first reported these details. He emphasized that C-suite technology and security leaders must collaborate to establish guardrails and concrete tracking for all agent tasks. "'The agent made me do it' is not going to be a defense from a moral or legal perspective."

Centralized platforms emerge as the dominant approach

Cisco took an aggressive stance when it launched its MyAgent platform in August, consolidating all company-authorized large language models, agents, and enterprise data into a single system. Thimaya Subaiya, executive vice president of operations at Cisco Systems, said the company will not authorize any third-party AI agents to prevent "agent sprawl" across departments.

The networking equipment maker built MyAgent on its own compute, storage, networking, security, and observability infrastructure. Around 90,000 employees gained access to the platform, which saw 50% daily adoption within two weeks. Employees can create their own agents, but a centralized team must approve them—roughly 700 have received authorization so far.

Intuit embedded security and risk controls into its generative AI operating system, GenOS, from the start. Chief Technology Officer Alex Balazs said the financial software company sketched "GenSRF"—representing security, risk, and fraud—alongside the initial architecture to ensure every AI request and response is tracked. "You don't want to try to retrofit the ability to enforce security and responsible AI foundations after the fact," Balazs noted.

New identity management systems for digital workers

Workday created an "agent system of record" to manage all non-human identities in its digital workforce, using the system internally and selling it to customers. CTO Gabe Monroy stressed that training is essential—engineers and other users need to understand both the risk profile and value proposition of each agent.

ServiceNow deployed a similar approach with its AI Control Tower, a platform to manage, observe, secure, and govern AI agents. Amit Zavery, the company's president, chief product officer, and chief operating officer, said the product is "probably one of the fastest-growing products ServiceNow has ever built" because it provides reassurance to executives and boards.

Why it matters

The shift from supervised AI tools to autonomous agents represents a fundamental change in enterprise risk management. Unlike traditional software, AI agents can take initiative and behave non-deterministically, creating what Sam Curry, chief information security officer at Zscaler, called "a new form of insider." Security professionals who spent careers managing human risk now face silicon-based intelligence with less understood motivations and incentive structures. Organizations that fail to establish robust governance frameworks before widespread agent deployment may face liability issues, security breaches, or operational failures that cannot be attributed to human error.

Jim Fowler, chief technology and product officer at Luman Technologies, argued against slowing adoption despite governance challenges. "I think for the broader enterprise market, the answer is secure acceleration, not slowing down," he said. "The bad guys aren't going to slow down, other nations aren't going to slow down."

Zscaler recently joined the Open Secure AI Alliance—a coalition led by Nvidia that includes Cisco Systems and Workday—focused on developing open-source tools with proper safeguards for software and AI agents.

These developments were first reported by Fortune.