Uncover how Cisco Buyer Expertise (CX) leveraged AI-ready infrastructure—together with networking, compute, and observability—to safe delicate information, management prices, and maximize ROI on agentic AI workloads.
AI isn’t simply shifting; it’s sprinting.
By strategically leveraging AI, we create proactive, personalised, and predictive buyer experiences that improve satisfaction and loyalty — with out sacrificing safety or price range. As capabilities quickly evolve, Cisco CX is remodeling a vital a part of its backend infrastructure to help superior AI workloads.
Tackling safety and price challenges
Though cloud platforms excel at flexibility and pace, they introduce some information sovereignty challenges and consumption-based value volatility. For AI workloads processing delicate buyer data, we determined that trade-off was untenable.
Safety: Cloud environments could distribute information throughout a number of networks and even geographically disperse areas, with entry controls by distinct third events — increasing the assault floor in an intensifying risk panorama. On-premises deployment eliminates these intermediaries. Cisco infrastructure met our information sovereignty wants, fuses safety throughout each layer of the stack, and reduces publicity to distributed threats.
Price: On-premises deployment additionally grants larger management over operational bills. Within the cloud, AI use instances with unpredictable inferencing calls for trigger token prices to skyrocket. For instance, a chatbot dealing with buyer inquiries could course of 10 million tokens throughout peak season however solely 2 million throughout off-peak intervals — leading to 5x value variance month-to-month. This unpredictability makes budgeting troublesome and compounds as AI workloads scale. After we course of tokens domestically, we change that ingress and egress volatility with value certainty, changing variable bills into predictable capital investments.
On-premises infrastructure mitigates two vital enterprise dangers: information sovereignty loss and price range unpredictability. By eliminating third-party intermediaries and variable token prices, we acquireed each digital resilience and monetary predictability to responsibly scale AI.
Deploying AI workloads flexibly
Since every AI use case has distinct necessities for information safety and computational scale, a inflexible “one-size-fits-all” deployment technique undermined our safety and price goals. As an alternative, we guided our selections utilizing pre-defined standards for AI use case prioritization, funding, and deployment. The three key goals had been:
- Good and scalable: Maximize the worth of buyer investments in Cisco applied sciences and options.
- Customizable expertise: Tailor a proactive, predictive, and personalised journey.
- Digital resilience and safety: Create a resilient, dependable, and safe buyer atmosphere.
These goals formed the place we deployed every workload, both on-premises, within the cloud, or hybrid. Take our Buyer Sentiment Evaluation Agent for example. It analyzes alerts to drive buyer renewals by processing a large scope of delicate information from buyer adoption journeys, help interactions, and the Cisco set up base. Due to its information sensitivity and scale necessities, on-premises deployment was each the safe and cost-effective selection — permitting us to take care of full management over buyer renewal information whereas avoiding unpredictable token prices throughout peak evaluation intervals.
With the help from this and different brokers, Cisco CX had 30% higher accessibility to adoption metrics versus guide assessments and eradicated day by day administrative friction as much as 40%.
Harnessing Cisco compute and networking
To scale AI workloads whereas sustaining information sovereignty and operational value predictability, we leveraged the next Cisco elements:
- Cisco Unified Computing System (UCS) Servers deal with the compute calls for of AI workloads, corresponding to mannequin tuning, utility inferencing, and job automation. The unified structure simplifies scaling and administration, enabling our workforce to develop AI capabilities with out the budgeting uncertainty that accompanies cloud-based inferencing.
- Cisco Nexus 9000 Collection Switches with Silicon One ASICs (Utility-Particular Built-in Circuits) present the low-latency, high-throughput networking required for intensive AI operations. Their programmable design reduces operational overhead throughout scaling occasions, making certain our infrastructure can adapt to workload calls for with out introducing new safety vectors or complexity.
- Splunk Cloud Platform delivers real-time visibility and infrastructure well being throughout your complete stack. This visibility is crucial for sustaining safety posture and operational effectivity — so we will successfully detect anomalies, optimize useful resource utilization, and guarantee predictable efficiency as workloads scale.
Finest practices and learnings
By deploying Cisco compute, networking, and observability options in tandem with CX agentic capabilities, Cisco CX ensures the end-to-end buyer lifecycle stays safe, seamless, and cost-effective. As we proceed to scale AI workloads, right here’s what we’ve discovered:
- Align on highest worth use instances: Prioritization isn’t subjective. Set up clear standards to guage the worth of use instances and deploy accordingly, so that you don’t should compromise safety or value.
- Prioritize reusability of AI infrastructure: Design your AI infrastructure as a shared platform, not siloed assets. The on-premises cluster that powers our Renewals Brokers additionally helps CiscoIQ, eliminating redundant {hardware} investments and accelerating time-to-value for brand new agentic workflows. This “construct as soon as, deploy many” method maximizes ROI and allows speedy scaling with out proportional infrastructure prices.
- Embrace steady analysis of your deployment mannequin: Simply as your infrastructure wants flexibility, your groups should commonly assess and adapt processes to optimize efficiency and price. Acknowledge that high-value use instances evolve with market situations and buyer wants — your infrastructure technique ought to too.
- Speed up time-to-market: Design your infrastructure for reusability and adaptability to cut back deployment cycles for brand new AI workflows. As an alternative of constructing customized infrastructure for every use case, groups can rapidly provision new workloads, creating extra time for experimentation.
By investing in the appropriate infrastructure and mindset, Cisco CX created house for each our workforce and prospects to innovate and thrive.
Associated Hyperlinks:
