Cisco President: AI Agents Need Reliability Checks Just Like Employees

Арестова Татьяна Exclusive
VK X OK WhatsApp Telegram
Cisco President: AI agents need reliability checks, just like employees

At the recently held AI summit in Amsterdam, Patel announced that Cisco has already developed a product whose code is entirely generated by AI. “By the end of 2026, we plan to have at least half a dozen products developed exclusively with the help of AI,” he noted.

In response to a question about the future of programmers, Patel advised not to fear competition from AI: “Instead, be wary of those who can use AI better than you. There will be no programmers at Cisco who do not master working with AI as a primary focus,” he said.

According to the president, the transition from traditional agile development to a specification-driven model will allow the development team to shrink from eight to three people, while adding five AI agents to significantly increase productivity.

AI agents play a key role in this process, as they are capable of planning and executing tasks with minimal human intervention. For example, such agents can handle preparing and sending emails, as well as organizing trips.

However, Patel noted that traditional developers will still need to check and edit the code.

He also questioned the traditional concept of human-in-the-loop, which implies mandatory human involvement in AI operations. Instead, he suggested viewing AI as an integrated participant in every stage of work, rather than just a tool.

Issues with AI agent control

Despite his optimistic view on AI, Patel expressed concerns about the safety associated with AI agents. “The most concerning issues for me remain those of security. It is important not to ignore the potential negative consequences of this technology,” he emphasized.

“These agents, performing tasks on our behalf, must undergo reliability checks, just like any other employees,” he added.

According to Patel, Cisco plans to invest billions of dollars in security to address two key challenges: to protect AI agents from external threats and to prevent potential risks associated with their uncontrolled behavior.

“We must protect the agent from the world, and the world from the agent,” he stressed.

The company aims to become a crucial part of the infrastructure of the new AI era, creating what is referred to as a “full stack,” encompassing everything from silicon chips to application solutions.

By developing everything from specialized chips to software, Cisco aims to become a critically important part of the infrastructure needed in the AI era.

“When we develop hardware and software simultaneously, real magic happens, as everything works in harmony,” he explained.

Speaking about the future, Patel noted that even significant productivity growth is not the main goal. “The main thing is the new ideas and knowledge that AI can generate, which are absent in the current array of human knowledge, allowing us to solve problems we never even dreamed of,” he said, mentioning potential breakthroughs in healthcare, poverty alleviation, and other important areas.
VK X OK WhatsApp Telegram

Read also: