Discover how Cisco transformed customer support into a context-aware, in-product experience that now reaches more than 250,000 users across our core portfolio. By leveraging a decade of operational experience and roughly 1.7 million annual customer cases, we have unified intelligence across cloud, on-premises, and sovereign cloud environments to deliver proactive guidance and seamless escalation.
Over the past several years as a Principal Engineer at Cisco, I have helped architect Cisco’s in-product experience, designing how context-aware guidance and AI-driven support get delivered directly inside the customer’s workflow. My perspective is shaped by a unique vantage point: I have also served as a technical judge for more than 500 AI, security, and enterprise technology submissions across industry awards, hackathons, and standards forums. This blog explores what we built at Cisco and how evaluating systems across the industry has confirmed our approach.
Our mission: improved in-product experience
Customer experience is always a top priority at Cisco. We noticed a recurring issue we call the “three-F problem:” fatigue, friction, and frustration. When problems occurred, customers had to leave the product, switch between multiple portals, repeat their context, and piece together help from different sources.
To address this, we created a unified experience layer that scales across cloud, on-premises, and sovereign cloud environments. This layer delivers contextual tooltips, inline guidance, banners, proactive alerts, guided walkthroughs, and an embedded AI assistant that can call specialized sub-agents and trigger evidence capture without leaving the product. This foundation led to Cisco’s in-product experience and AI-driven support, with smooth escalation to human engineers wherever customers are. The same intelligence appears across product UIs, Cisco.com, and support workflows, without requiring every team to rebuild the same scaffolding.
To deliver consistent experiences, intelligence needs to be accessible across product UIs, Cisco.com, and support workflows, across cloud, on-premises, and sovereign cloud environments. It is a unified capability that few other vendors offer at this scale. It is also grounded in more than a decade of technical support case data, with roughly 1.7 million customer cases flowing through Cisco’s support organization every year. That data is operationalized rather than archived, feeding directly into what the AI surfaces in real time, inside the product, at the moment a customer needs it.
Using external insights to inform internal strategy
Cisco’s own operational experience drove the architectural decisions described in the rest of this article. The external signals discussed below are not the source of those decisions. They are an independent check on them.
1. Intelligence in isolation vs. real world context
Our early experiments revealed a recurring problem: systems showed strong intelligence in isolation, through advanced models and polished demos, but struggled in real-world scenarios because they lacked understanding of user intent, system status, operational limits, and lifecycle context.
Reviewing hundreds of external submissions later showed the same pattern at industry scale, which confirmed the architectural direction we had already taken.
This reinforced a core architectural decision at Cisco: intelligence alone would not fix the experience. Context had to be treated as a first-class concern.
The solution: Unifying operational view
This conviction shaped how we designed our in-product experience to operate as a continuous, context-aware system. We unified telemetry, user behavior, and product signals into a shared operational view.
When an issue arises, the system proactively offers guidance, explains the impact, answers questions, and helps fix the problem, all within the product.
2. Scaling AI support across enterprise workflows
Our early internal experiments showed that a single monolithic agent struggled to maintain accuracy, explainability, and trust as workflows became more complex. Reviewing external submissions confirmed the same pattern industry-wide: there is no single, generic AI assistant able to work across enterprise workflows.
The solution:
External signals validated our belief: autonomy without structure doesn’t scale. So, we built specialized agents that work together, directing users to the right expert instead of relying on one assistant to handle everything. This allows the system to progress through investigations step by step, preserve context across agents, and hand off cleanly when human expertise was required.
Today, the system is delivered as a set of specialized AI agents working together rather than competing for control. At the start of the workflow, a Case Management Agent gathers the support bundle, logs, and version data without the customer needing to attach them manually. A Configuration Analysis Agent scans the configuration for invalid statements, rule conflicts, and version-dependent mismatches. A PSIRT and Field Notice Agent evaluate known vulnerabilities and advisory applicability against the customer’s environment and entitled devices. When the workflow needs to escalate, the system delivers a TAC-ready case with logs, configs, and reasoning trace already attached, so the engineer on the other end starts with full context.
3. Designing for operations, not just demos
Inside Cisco, we prioritized stability, reuse, and consistency over rapid feature sprawl from the start. External evaluations later highlighted the same divide we had been navigating between systems built for demonstration and those designed for reliable operations.
The solution:
By focusing on operational behavior from the start, we provided guided workflows across products and portals that now lead users through fixes and send proactive alerts.
For engineering, this reduced duplication; hundreds of previously fragmented workflows now run through a common experience layer, saving development effort and ensuring consistency.
The result was not just a better customer experience, but a more sustainable way to scale innovation internally.
The impact has been measurable. More than 25,000 customer workflows are delivered through in-product self-service every week, with resolution times 25 to 30% faster than traditional support paths. For engineering, this consolidation has streamlined hundreds of previously fragmented workflows and given product teams a unified foundation they no longer have to rebuild per release.
Unexpected wins
Customers shared, “At last, I can show you what’s happening instead of struggling to explain it.”
The unexpected insight was not that simple solutions win, but that customers most wanted a low-friction way to transfer context, to show rather than describe, to the human on the other end. Screen recording happened to be the most direct mechanism that delivered that. The lesson generalized: the highest-value AI capability is often whatever cleanly captures and transfers context between humans, enhancing communication on both sides rather than replacing it.
The most valuable thing AI can do is not always the most complex – it’s often about capturing context cleanly so humans on either end can communicate effectively.
Takeaways and looking ahead
The value of external evaluation is not validation — it’s calibration. Reviewing hundreds of external systems sharpened our understanding of which design choices hold up under pressure and which quietly fail over time.
Today, our in-product capabilities reach more than 250,000 users across Secure Firewall, Wireless LAN Controller (WLC), Cisco XDR, SD-WAN, Secure Access, Email Threat Defense, and Cisco.com and support portals. The patterns that informed this approach have also been recognized externally: a Gold Stevie, three Silver Stevies, the 2026 Edison Gold Award, the 2026 CODiE Award for Best Knowledge Management and Search Solution, and the ISSIP Impact to Business Service Innovation Award for the in-product experience platform.
By keeping one foot in external evaluation and the other in internal execution, we’ve moved from reactive support to proactive, in-product experiences that scale with the business.
Resources
