Discover how Cisco Customer Experience (CX) leveraged AI-ready infrastructure—including networking, compute, and observability—to secure sensitive data, control costs, and maximize ROI on agentic AI workloads.
AI isn’t just moving; it’s sprinting.
By strategically leveraging AI, we create proactive, personalized, and predictive customer experiences that enhance satisfaction and loyalty — without sacrificing security or budget. As capabilities rapidly evolve, Cisco CX is transforming a critical part of its backend infrastructure to support advanced AI workloads.
Tackling security and cost challenges
Although cloud platforms excel at flexibility and speed, they introduce some data sovereignty challenges and consumption-based cost volatility. For AI workloads processing sensitive customer information, we decided that trade-off was untenable.
Security: Cloud environments may distribute data across multiple networks and even geographically disperse locations, with access controls through distinct third parties — expanding the attack surface in an intensifying threat landscape. On-premises deployment eliminates these intermediaries. Cisco infrastructure met our data sovereignty needs, fuses security across every layer of the stack, and reduces exposure to distributed threats.
Cost: On-premises deployment also grants greater control over operational expenses. In the cloud, AI use cases with unpredictable inferencing demands cause token costs to skyrocket. For example, a chatbot handling customer inquiries may process 10 million tokens during peak season but only 2 million during off-peak periods — resulting in 5x cost variance month-to-month. This unpredictability makes budgeting difficult and compounds as AI workloads scale. When we process tokens locally, we replace that ingress and egress volatility with cost certainty, converting variable expenses into predictable capital investments.
On-premises infrastructure mitigates two critical business risks: data sovereignty loss and budget unpredictability. By eliminating third-party intermediaries and variable token costs, we gained both digital resilience and financial predictability to responsibly scale AI.
Deploying AI workloads flexibly
Since each AI use case has distinct requirements for data security and computational scale, a rigid “one-size-fits-all” deployment strategy undermined our security and cost objectives. Instead, we guided our decisions using pre-defined criteria for AI use case prioritization, investment, and deployment. The three key objectives were:
- Smart and scalable: Maximize the value of customer investments in Cisco technologies and solutions.
- Customizable experience: Tailor a proactive, predictive, and personalized journey.
- Digital resilience and security: Create a resilient, reliable, and secure customer environment.
These objectives shaped where we deployed each workload, either on-premises, in the cloud, or hybrid. Take our Customer Sentiment Analysis Agent as an example. It analyzes signals to drive customer renewals by processing a wide scope of sensitive data from customer adoption journeys, support interactions, and the Cisco install base. Because of its data sensitivity and scale requirements, on-premises deployment was both the secure and cost-effective choice — allowing us to maintain full control over customer renewal data while avoiding unpredictable token costs during peak analysis periods.
With the support from this and other agents, Cisco CX had 30% better accessibility to adoption metrics versus manual assessments and eliminated daily administrative friction up to 40%.
Harnessing Cisco compute and networking
To scale AI workloads while maintaining data sovereignty and operational cost predictability, we leveraged the following Cisco components:
- Cisco Unified Computing System (UCS) Servers handle the compute demands of AI workloads, such as model tuning, application inferencing, and task automation. The unified architecture simplifies scaling and management, enabling our team to grow AI capabilities without the budgeting uncertainty that accompanies cloud-based inferencing.
- Cisco Nexus 9000 Series Switches with Silicon One ASICs (Application-Specific Integrated Circuits) provide the low-latency, high-throughput networking required for intensive AI operations. Their programmable design reduces operational overhead during scaling events, ensuring our infrastructure can adapt to workload demands without introducing new security vectors or complexity.
- Splunk Cloud Platform delivers real-time visibility and infrastructure health across the entire stack. This visibility is essential for maintaining security posture and operational efficiency — so we can effectively detect anomalies, optimize resource utilization, and ensure predictable performance as workloads scale.
Best practices and learnings
By deploying Cisco compute, networking, and observability solutions in tandem with CX agentic capabilities, Cisco CX ensures the end-to-end customer lifecycle remains secure, seamless, and cost-effective. As we continue to scale AI workloads, here’s what we’ve learned:
- Align on highest value use cases: Prioritization isn’t subjective. Establish clear criteria to evaluate the value of use cases and deploy accordingly, so you don’t have to compromise security or cost.
- Prioritize reusability of AI infrastructure: Design your AI infrastructure as a shared platform, not siloed resources. The on-premises cluster that powers our Renewals Agents also supports CiscoIQ, eliminating redundant hardware investments and accelerating time-to-value for new agentic workflows. This “build once, deploy many” approach maximizes ROI and enables rapid scaling without proportional infrastructure costs.
- Embrace continuous evaluation of your deployment model: Just as your infrastructure needs flexibility, your teams must regularly assess and adapt processes to optimize performance and cost. Recognize that high-value use cases evolve with market conditions and customer needs — your infrastructure strategy should too.
- Accelerate time-to-market: Design your infrastructure for reusability and flexibility to reduce deployment cycles for new AI workflows. Instead of building custom infrastructure for each use case, teams can quickly provision new workloads, creating more time for experimentation.
By investing in the right infrastructure and mindset, Cisco CX created space for both our team and customers to innovate and thrive.
Related Links:
