Cisco is doing something most business leaders talk about but haven’t actually pulled off. Starting in its new fiscal year at the end of July 2026, every one of the company’s 90,000 employees gets a personalised AI agent.

Not a shared chatbot. Not a pilot program for one department. Every single person.

The announcement, detailed by Cisco CFO Mark Patterson in Fortune, is one of the largest enterprise AI deployments ever attempted. And the practical details of how Cisco is doing it are more instructive than the headline number.

It’s Not One Model. It’s a Routing Layer.

The most interesting technical detail isn’t the scale. It’s the architecture. Cisco’s system doesn’t default to the most powerful frontier model for every task. Instead, it routes each request to whichever model is most cost-efficient for that specific job.

A simple query about company policy goes to a lightweight model. A complex financial analysis goes somewhere more capable. Most of the infrastructure runs on-premises, giving Cisco direct control over both costs and data security.

This is the opposite of how most organisations are deploying AI right now. The default assumption has been: use the best model available and deal with the bill later. Cisco is running it like an operations problem from day one.

Finance Is the Test Case

Patterson’s own team is the proving ground. In Cisco’s finance function, AI already produces 80 to 90 percent of the first draft of the MD&A section in public regulatory filings. What would have taken analysts days now takes minutes, with humans reviewing and refining rather than starting from scratch.

He’s now building what he calls a “CFO cockpit” dashboard. The idea: a single interface that synthesises performance data from across the business and surfaces recommended actions. Less reporting, more deciding.

That shift from reporting to decision-support is the pattern most finance and operations leaders are trying to replicate. Cisco is further along than most.

The Layoff Question

The timing is uncomfortable. Cisco told staff in May 2026 that it would cut fewer than 4,000 jobs, under 5 percent of its global workforce, as part of a restructuring to invest more heavily in AI infrastructure.

So the company is simultaneously giving AI agents to 90,000 people while removing 4,000 roles. That tension is real, and it’s the same tension playing out at dozens of enterprises this year. The honest answer is that AI deployment and workforce restructuring are happening at the same time because they’re connected. Tasks that AI can handle reliably are being automated. People are being redirected toward work that still requires human judgement.

What Cisco is doing differently is making that trade explicit rather than quiet. They’re announcing the agent deployment and the job cuts in the same breath rather than hoping nobody notices the correlation.

What This Means for Business

Cisco’s playbook contains several moves that business leaders outside of tech can adapt right now.

Start with finance. Patterson chose his own function as the test case. That gives him direct visibility into what’s working, forces the team to use the tools they’re supposed to be championing, and produces measurable ROI quickly. CFOs and COOs who lead AI adoption rather than waiting for IT to deliver it are moving faster.

Build the routing layer. Most organisations are paying frontier model prices for tasks that don’t need frontier model capability. A routing architecture that matches task complexity to model cost is a legitimate competitive advantage as AI spend scales.

Run it on-premises where it matters. Cloud-first AI makes sense for many workloads. But for sensitive data and cost-sensitive, high-volume operations, on-premises deployment gives you control that pure cloud doesn’t. Cisco’s infrastructure choice reflects a mature operational stance.

Measure in decisions, not outputs. The CFO cockpit framing is the right mental model. The goal isn’t to produce more reports faster. It’s to make better decisions with less manual effort. That distinction shapes what you build.

Enterprise DNA works with business leaders who are navigating exactly this transition. Whether you’re building your first AI agent workflow or trying to scale what’s already working, the Cisco approach offers a useful framework: start with a real problem, build the routing layer, measure in business outcomes, and don’t pretend the workforce implications aren’t there.

The companies getting ahead right now are the ones treating AI deployment as an operations challenge, not a technology experiment.

---

Want the practical version of this? The free Working With Claude field guide covers the full Claude ecosystem, Claude Code, and how to roll it out across a real business. Download it here.