Cisco has deployed MyAgent, a personalized AI agent, to all 90,000 of its employees, according to a company blog post by Thimaya Subaiya, EVP Operations at Cisco, published on 27 August 2026. The rollout began in late July 2026 and runs on Circuit, Cisco’s internal, multi-model AI platform. MyAgent executes supervised, multi-step workflows across Outlook, Webex, Jira, and SharePoint from a single employee-stated objective, making the deployment one of the largest corporate AI-assistant rollouts recorded to date.
Cisco built MyAgent to control cost as much as to add capability. According to architecture figures published by Cisco and reported by AI researcher Rohan Paul on X, the system routes 50 to 60 percent of requests to open-weight models, 20 to 30 percent to software automation, and only a small remainder to frontier models.
- Scale: 90,000 employees, every Cisco worker worldwide
- Platform: Circuit, Cisco’s secure, governed, model-agnostic AI platform
- Architecture: more than 800 specialized subagents behind one personal agent
- Integrations: Outlook, Webex, Jira, and SharePoint
- Cost routing: 50–60% open-weight models, 20–30% software automation, remainder frontier models
How Cisco Built MyAgent on the Circuit Platform
MyAgent is a personal orchestration layer that sits on Circuit and dispatches work to more than 800 specialized subagents, each handling one task type across the connected systems: Outlook, Webex, Jira, and SharePoint. An employee states an objective, and MyAgent plans the steps, calls the relevant subagents, and completes the workflow under human supervision.
Persistent memory separates MyAgent from a chatbot. MyAgent retains user preferences, past interactions, and working context across sessions, so a request made in September builds on decisions recorded in August. Cisco describes the result as “ambient intelligence”: Subaiya’s post is titled “MyAgent and the Rise of Ambient Intelligence: Cisco’s Next Step in Enterprise AI,” and the Cisco blog post summarizes the goal as delivering “secure, personalized ambient intelligence with persistent context that turns intent into action.”
How Cisco Controls AI Agent Costs Across 90,000 Users
Cisco controls cost with a three-tier routing pyramid: open-weight models handle 50 to 60 percent of requests, deterministic software automation handles 20 to 30 percent, and frontier models handle only the small remainder that requires top-tier reasoning. The design keeps the most expensive model calls to a minority of traffic at a scale of 90,000 users.
The pyramid inverts the default enterprise pattern, in which every prompt goes to a single frontier model. Cisco escalates upward only when cheaper tiers fail: software automation for routine steps, open-weight models for the majority of model calls, and frontier models for the small remainder. Cisco has not published the resulting per-employee cost.
What Cisco’s Rollout Means for Businesses Choosing AI Agents
Cisco’s deployment is the clearest public example of an enterprise scaling AI agents beyond a pilot, a shift McKinsey’s 2026 State of AI survey quantifies: enterprises scaling agents in at least one function rose from 27 percent to 40 percent year on year, while smaller firms stayed flat at 22 percent. Business buyers evaluating agent platforms now have a reference architecture to measure vendors against. Cisco’s checklist of model-agnostic routing, cross-system integrations, persistent memory, and supervised autonomy is the same set of criteria we apply in our comparison of the 12 best AI agents for business tasks, where vendors such as Microsoft Copilot Studio, Zapier AI Agents, and Claude Code are scored on governance and per-task cost.
The Cisco story is an operations and IT case. Salesforce’s Claudeforce partnership, announced on 26 August 2026, is the CRM equivalent: a different buyer, the same trend toward agents that act across systems rather than answer inside one.
What Cisco Has Not Disclosed
Cisco has not published productivity gains, return-on-investment figures, or cost savings against previous tools. Cisco has not named the frontier models behind the top routing tier. Cisco has not stated whether MyAgent or Circuit will ship as a commercial product; every available source, including PYMNTS’ report on the deployment, describes MyAgent as an internal deployment.
For Context
Enterprise agent deployments accelerated through August 2026. Salesforce and Anthropic launched Claudeforce with 37 AI skills for CRM one day before Cisco’s blog post, and Temporal’s 2026 report found daily AI agent use jumped 70.8 percent in a year. Cost control is the recurring constraint: Google added spend caps to Gemini Enterprise after AI sticker shock, the same problem Cisco’s routing pyramid is built to prevent.
Related Coverage
- Best AI Orchestration Tools 2026: 10 Platforms for Multi-Agent Workflows Compared — the commercial platforms that coordinate subagents the way Circuit coordinates MyAgent’s 800
- Varonis Can Now Block AI Agents That Go Off-Script in Enterprise — the governance layer a 90,000-user agent deployment depends on
- Nvidia, Microsoft & Meta: Don’t Restrict Open-Weight AI — why the open-weight models carrying 50–60% of Cisco’s traffic matter to policy
FAQ
What is Cisco MyAgent?
Cisco MyAgent is a personalized AI agent deployed to all 90,000 Cisco employees, built on Cisco’s internal Circuit platform. MyAgent takes an employee’s stated objective and completes multi-step workflows across Outlook, Webex, Jira, and SharePoint under human supervision, using more than 800 specialized subagents.
How does Cisco keep AI agent costs down?
Cisco routes 50 to 60 percent of MyAgent requests to open-weight models and 20 to 30 percent to software automation. Only the small remainder reaches expensive frontier models, which limits the highest-cost calls to a minority of traffic.
Can businesses buy Cisco MyAgent?
No. Cisco has not announced MyAgent or Circuit as a commercial product, and every published source describes the deployment as internal to Cisco. Businesses can apply the same architecture using commercial agent platforms with model routing and cross-system integrations.
How many companies are scaling AI agents in 2026?
McKinsey’s 2026 State of AI survey reports that 40 percent of enterprises are scaling agents in at least one function, up from 27 percent a year earlier. Smaller firms stayed flat at 22 percent over the same period.
