The networking giant has given roughly 90,000 workers access to personalized AI agents, turning one of the world’s largest technology companies into a live experiment in how agentic AI could reshape everyday corporate work.
Cisco has begun giving each of its roughly 90,000 employees access to a personalized artificial-intelligence agent, marking one of the largest company-wide deployments yet of AI designed not merely to answer questions, but to carry out work on behalf of employees.
The system, known internally as MyAgent, is intended to function as a digital workplace assistant capable of managing emails, analyzing information, monitoring market developments and helping employees complete a wide range of routine and analytical tasks. Workers can access it through desktop and mobile devices as well as through Cisco’s Webex collaboration platform.
The rollout represents a significant step beyond the corporate chatbot experiments that have proliferated across the technology sector over the past several years.
Traditional generative-AI assistants typically wait for a user to ask a question and then generate a response. Agentic systems are designed to go further: they can break a request into multiple steps, select tools, retrieve information, interact with internal systems and complete portions of a workflow with limited human intervention.
For Cisco, that distinction is central to the project.
Each employee’s agent can call on a network of more than 800 specialized subagents, according to reporting by The Wall Street Journal. These smaller systems are designed to perform particular tasks or interact with specific corporate functions, while the employee’s primary agent acts as an orchestrator.
The system is personalized according to the individual employee’s permissions, meaning an agent should only be able to access the information and internal resources that its human user is authorized to see.
That architecture is important because one of the largest barriers to enterprise AI adoption remains security. Companies increasingly want employees to use powerful AI tools, but they also need to prevent confidential information, customer data and proprietary intellectual property from leaking into uncontrolled public systems.
Cisco has therefore built security controls directly into its internal AI environment, including policy mechanisms that govern access to company information and monitor how AI tools are being used.
The company has been experimenting with internal AI assistants for some time. Cisco said last year that users of its earlier internal assistant reported meaningful productivity improvements, with 73 percent saying it increased their productivity and average reported time savings of about five hours per week.
The new agent rollout takes that strategy considerably further.
Rather than limiting advanced AI systems to engineers, finance teams or selected experimental groups, Cisco is attempting to make agentic AI part of the everyday working environment across the entire organization.
Chief Financial Officer Mark Patterson has described AI as one of the most consequential technological transitions of his career. He said Cisco’s aim is not simply to give employees access to the largest or most expensive AI model available, but to build an architecture that automatically routes each task to the model best suited to perform it.
That approach addresses another increasingly important problem in corporate AI: cost.
Agentic systems can consume far more computing resources than conventional chatbots because they may perform numerous intermediate calculations, queries and tool calls before completing a task. At the scale of tens of thousands of employees, those costs can rise rapidly.
Cisco is therefore using a mixture of internally operated infrastructure, open-weight AI models and commercial foundation models. The company can direct relatively simple tasks toward cheaper models while reserving more powerful systems for requests that actually require them. According to The Wall Street Journal, roughly 50 to 60 percent of Cisco’s AI activity currently runs on open-weight models, with only a smaller portion routed to more expensive frontier systems.
Much of the underlying infrastructure also runs on Cisco-controlled computing resources rather than depending entirely on external cloud providers.
Patterson has argued that this gives the company greater control over both data and operating costs.
The finance department already offers an indication of how dramatically these tools could alter corporate workflows.
Cisco says AI is now responsible for roughly 80 to 90 percent of the first draft of its management discussion and analysis material used in financial reporting. The company has also developed AI tools capable of examining previous earnings calls, comparing Cisco with competitors and anticipating questions that individual financial analysts might ask executives.
Patterson himself uses an AI agent for benchmarking, allowing him to compare Cisco’s revenue growth, research spending, earnings and capital allocation with competitors more quickly.
Other departments are expected to discover their own applications.
Sales teams could use agents to prepare customer briefings or analyze account histories. Engineers could use them to search technical documentation and troubleshoot systems. Legal departments could review contracts and regulatory materials, while customer-support teams could allow agents to retrieve information and coordinate responses across multiple internal databases.
That breadth makes Cisco’s deployment particularly important for the wider technology industry.
For years, companies have talked about AI as a productivity tool. Agentic AI raises a more consequential possibility: that software may increasingly become an active participant in corporate workflows rather than simply a tool employees consult.
Cisco itself describes the emerging workplace as one in which agents can plan tasks, invoke software tools, access systems and trigger workflows alongside human employees.
But the transformation also raises difficult questions about employment.
Cisco’s agent rollout comes during a period in which technology companies have been restructuring workforces while simultaneously investing heavily in artificial intelligence. That juxtaposition has intensified employee concerns that productivity tools presented initially as assistants may eventually reduce demand for some categories of work.
The challenge for corporate leaders will therefore be not only technical, but cultural.
Employees will need to trust the agents enough to use them while also understanding where human oversight remains necessary. Companies will have to determine which decisions can safely be delegated, who remains accountable when an automated system makes a mistake and how performance should be measured when increasingly large portions of work are completed by machines.
There is also the question of skills.
Cisco has paired its AI deployment with internal training and knowledge-sharing initiatives intended to encourage employees to experiment with the technology and identify new uses for it. Patterson has suggested that departments could effectively compete to discover the most productive applications.
That approach reflects a broader assumption taking hold across Silicon Valley: the future workplace may not be divided simply between jobs performed by humans and jobs performed by AI.
Instead, competitive advantage could increasingly depend on how effectively individual workers learn to direct, supervise and collaborate with autonomous software.
Cisco has considerable commercial incentive to prove that model works.
The company is positioning itself as one of the major infrastructure providers for the AI economy, selling networking equipment, computing systems, security technology and custom silicon needed to operate increasingly large AI workloads. Cisco reported $9.3 billion in AI infrastructure orders from hyperscale customers during fiscal 2026, underscoring how important the sector has become to its growth strategy.
Just this week, Cisco also expanded its Secure AI Factory initiative with Nvidia and Supermicro, targeting enterprises and cloud providers seeking high-density computing infrastructure capable of running enormous AI models while maintaining control over data and operating costs.
Deploying agents across its own workforce therefore serves a dual purpose.
Cisco is not merely adopting the technology. It is also using itself as a demonstration of what an AI-intensive enterprise might eventually look like.
If the experiment succeeds, the company could provide an influential blueprint for other large organizations contemplating similar deployments: one AI agent for every employee, backed by specialized subagents and connected securely to the systems that actually run the business.
The deeper significance of the initiative lies in that scale.
Giving a few hundred engineers access to advanced AI is an experiment. Giving roughly 90,000 workers their own digital agent begins to look like a new operating model.
Cisco is betting that the next phase of workplace AI will not be defined by employees occasionally opening a chatbot.
It will be defined by employees arriving at work with an artificial colleague already waiting for instructions.
