A Company Just Gave Every One of Its 90,000 Employees an AI Agent. Here's Why That's a Big Deal

Cisco just did something almost no company its size has actually pulled off. It gave a personal AI agent to every single one of its roughly 90,000 employees, worldwide, at once.

Not a test group. Not a slow rollout to one department while the rest of the company watches. Everyone.

What actually happened

Cisco calls the tool MyAgent. It runs on top of Circuit, the company's internal AI platform, and it works inside the tools employees already use every day: Outlook, Webex, Jira, SharePoint. An employee can hand it a goal instead of a single question, and the agent coordinates the steps needed to move that work forward.

That's a real shift from the chatbot model most companies are still stuck on. A chatbot answers what you type. MyAgent is built to take a task, work across systems, and get it partway or all the way done, with a human still checking the output.

A few things about how Cisco built this matter more than the headline number:

• It doesn't default to the fanciest model for every task. A simple policy question gets routed to a lightweight, cheap model. A complex financial analysis gets routed to something more capable. That routing decision alone protects the budget from ballooning the way a lot of company-wide AI rollouts do.
• Most of it runs on infrastructure Cisco controls directly, which gives them a tighter grip on cost and data security instead of handing everything to a third party by default.
• The rollout came with training and internal competition between teams, not just a login and a shrug. Cisco is actively trying to surface new uses instead of hoping employees stumble into value on their own.

Cisco's own finance team is already a proof point. According to the company, AI now produces most of the first draft of the mandatory narrative section in its public financial filings, work that used to take analysts days and now takes minutes, with people reviewing and refining instead of starting from a blank page.

Why this is bigger than one company's IT project

Enterprise AI has mostly lived in pilot purgatory. A department tries a tool, likes it, and then nothing happens for six months while legal, IT, and finance argue about governance. Cisco skipped that. They went from pilot to full company deployment, and they built the guardrails (model routing, human review, controlled data access) into the rollout itself instead of bolting them on after something went wrong.

That matters because it's a live example of the next phase of AI adoption. The first phase was "can this thing answer questions well." The second phase, the one we're in now, is "can this thing be trusted to actually do work inside a real company, at real scale, without creating a mess." Cisco is one of the first companies to show that the answer can be yes, if you design for it on purpose.

What it means for a growing business

You don't have 90,000 employees. You don't need to. The lesson isn't the size, it's the shape of the decision.

Here's what carries over to a business with 20 people or 200:

• Don't run every task through your most expensive AI tool. Match the tool to the job. Simple, repetitive stuff can run on something lean and cheap. Only send the complicated work to your best resource, whether that's a person or a model.
• A personal agent beats a shared chatbot. An assistant that knows a specific employee's role, their recurring tasks, and the systems they touch every day is worth more than a generic tool everyone shares and nobody customizes.
• Adoption doesn't happen by accident. Cisco paired the rollout with training and internal pressure to use it. If you hand your team a new AI tool and walk away, most of them will ignore it. The plan for getting people to actually use it matters as much as the tool itself.
• Keep a human checking the output. Cisco's system is supervised, not autonomous. That's not a limitation, that's the responsible version of this. Speed without a check is how mistakes get expensive fast.

The scale of what Cisco did isn't the point. The point is that a serious company just proved you can build AI into daily work, department by department, task by task, without it turning into chaos. That's the blueprint, shrunk down to whatever size your business actually is.

The takeaway

AI agents inside a real business aren't a future idea anymore. A major company just ran the experiment at full scale, in public, and the useful part isn't the number 90,000. It's the discipline underneath it: right tool for the task, humans still in the loop, and a real plan to get people using it. That's a formula any growing business can copy, at a size that actually fits.