AI agents should be treated like employees who need vetting before they’re trusted with real work, says Cisco’s president.
Jeetu Patel said in an interview with Euonews Next published on Wednesday that AI agents working on our behalf “need to get the background checks done,” just like human employees.
As companies race to deploy autonomous AI systems that can write code, handle workflows, and make decisions, security needs to evolve just as quickly, he said.
“You have to protect the agent from the world of attacks, and you have to protect the world from the agent that goes rogue,” he added.
Patel’s comments come as Cisco accelerates its own AI adoption. He predicts that by the end of 2026, the company will have “at least half a dozen products” written entirely with AI, with no human-written lines of code, he said.
“We won’t have developers at Cisco that don’t use AI as a core habit within the way that they get the job done,” he said.
“Don’t worry about AI taking your job, but worry about someone using AI better than you definitely taking your job,” he added.
AI agents taking over work
Tech leaders are going all in on AI agents.
StackBlitz CEO Eric Simons told Business Insider he wants more AI agents than human employees at his startup this year — a milestone that reflects a broader transformation sweeping the software industry.
AI can already write code, manage tasks, and coordinate across platforms. OpenClaw, a personal AI assistant that works across messaging apps, is one example of how agents can operate on a human’s behalf with minimal involvement.
“This is a crystal ball into the wildness of the inevitable future,” Simons told Business Insider in a report published Wednesday.
“Your AI agents will talk to other people’s agents on your behalf — negotiating pricing for a product you want to buy, checking restaurant availability, even arguing your political viewpoints,” he added.
Software stocks slid in recent weeks as investors grappled with the idea that AI may replace software.
Still, there are warning signs. Agentic systems can become fragile at scale. If hallucinations multiply as agents interact, small errors can cascade, said Nicolas Darveau-Garneau, a former Google executive and author of “Be a Sequoia, Not a Bonsai.”
“The only way we get massive productivity gains is the day we can trust that the AI isn’t hallucinating a lot,” Darveau-Garneau said.
