Cisco Systems’ Liz Centoni shares wisdom from her hard-fought efforts to guide the tech titan’s $15 billion services business into an AI-native company.
Cisco Systems, a giant in networking, cybersecurity, and AI, is leveraging a combination of agentic AI, traditional AI, and automation to transform its business from top to bottom.
Liz Centoni, EVP and chief customer experience officer, is at the helm of the company’s $15 billion services business, an organization of 20,000 employees responsible for professional services, customer success, technical support, and the renewals engine for software and services.
In 2024, she led the company’s $28 billion acquisition of Splunk. Today, the 26-year Cisco veteran is helping guide the multinational tech conglomerate’s transformation into an AI-native company in terms of internal operations and what it delivers to its customers.
At the heart of her vision are these governing ideas:
- Every workflow constrained by human availability is a workflow for redesign.
- The status quo is not an option. Customer expectations and demands will continue to grow. You can’t rest on your laurels because revenue is growing and margins look good, so you must continually consider and reimagine your processes and offerings.
- The quality we have today is the floor, not the ceiling. The capabilities of AI continue to evolve at a rapid pace.
- Leadership, not technology, is the limiting factor. Leaders need to find and support the leaders underneath them who can amplify efforts and bring people along with them.
Based on her experience, Centoni also has four pieces of advice to achieve success with agentic AI.
1. Change the cost equation of expertise
“We didn’t just start bolting on AI,” Centoni says. “We looked at the underlying workflows to understand how we could redo work.”
This was one of the first lessons Centoni and her team took to heart. When she took on her current position two years ago, she spent time looking through feedback from customers. One of the biggest pet peeves customers expressed was the frustration of the path leading to the right expert to solve a problem. That process frequently required handoffs from one engineer to another, and explaining the problem in detail each time.
The team’s first reaction was to build an agent that would speed up the process to find the right engineer to handle the problem. That, Centoni says, was a misstep.
“It just upset our customers more because it showed that we actually did five handoffs behind the scenes and it surfaced it to them faster,” Centoni says. “We just bolted AI onto the wrong thing and said, ‘automation.’”
Before even thinking about creating an agent, what they should’ve asked, Centoni says, was why have handoffs in the first place. It’s not enough to simply map a workflow and attempt to automate it because steps of that workflow may have been designed decades ago to address an issue an agent may not have or be able to cut through. It’s essential to understand the desired end result, and then design a workflow with current capabilities and constraints in mind, she adds.
“Users are an essential part of the technology stack,” Centoni says. “You have to bring in the subject matter experts and the users to outline the workflow, and you have to get close to the work. If you can’t, you have no right designing.”
An important element, she continues, is to carefully consider the boundaries between where an agent can do something autonomously and where it’s necessary to hand the process off to a human. Where can you use automation and AI with 90%-plus accuracy?
That gets you to start rethinking the workflow, Centoni says, since throughout most of Cisco’s history, the approach to customer engagements was to throw experts at a problem. The more complex the engagement, the more experts were needed. For customers, the priority was getting the job done correctly but as quickly as possible, yet the availability of expert engineers throttled Cisco’s ability to speed up the process.
“When I came into this business, we treated every customer we had like a bespoke model,” Centoni says. “But over the last 40-plus years, we’ve solved these problems tens of thousands of times.”
Cisco has taken 40 years of engagements to build a knowledge base and a series of playbooks that AI agents and human experts can use to help customers plan and execute deployments. Cisco can now take the information they have about a customer, including their KPIs for a project, and customize its value realization without needing to include a lot of human experts on top.
“You go into a new deployment with an expert agent that brings in all the prior experts and democratizes the knowledge base,” Centoni says. “The outcome for the customer is consistency, security, and better time to project completion.”
Still, human experts are always in the loop, and Centoni says the team would have more bandwidth for conversations with the customer on how to iterate and help them feel they get more than what they paid for.
2. Be more proactive and predictive
“Over the years, our focus has always been on how fast we can get to root cause analysis of customer issues,” Centoni says. “If it took me three hours before, can I get it down to 30 minutes.”
Helping customers solve issues faster is admirable, but AI can help Cisco be more proactive and predictive. Instead of just speed, the company is now leaning on AI to help it stop issues before they hit a customer.
“When we look at case histories, we can see clusters that happen over and over again,” she says. Cisco leverages decades of understanding what can go wrong as a knowledge base. Agents can now continually monitor customer environments for anything that might signal an issue is occurring, and recommend next best steps before it becomes critical.
“In 12 months, I’d like to be able to sit down with a customer and say you would’ve had 10 outages last year and we prevented eight of them,” Centoni says.
While the past year has been all about AI adoption and token maxing, companies and their boards are increasingly paying attention to the ROI of their AI efforts and demanding results. The key to keeping costs down and returns high, Centoni says, is to be intentional about models and tools.
“Before we even go down the AI path, we ask, ‘Can automation do this?’” she says. “If it’s purely rules-based, I’m doing automation. If it requires bringing multiple signals together to drive a prediction, then AI. In many cases, it’s not binary. We have a combination of things where we use old predictive AI models, automation, and some of the newer LLMs.”
When an agent is called for, pick the models and tools that best fit the use case, not the preferences of the designer. For instance, a use case might involve data sovereignty requirements in which data must stay on prem.
“If you look at our AI assistant for renewals, which brings together a renewals agent, a sentiment agent, a risk agent, and an install base agent, it’s actually based on Mistral because they were one of the few who could give us an on-prem model that had the full capabilities,” Centoni says. “We customized that model using a private instance and we use automation along with it as well.”
4. Avoid compliance theater
A lot of tech leaders measure their AI success by adoption. Not Centoni. If you measure adoption, a lot of users will just check the box to say they’ve used it, and then use the old processes and tools to get the actual work done. Centoni calls it compliance theater.
If she wants to understand the success of the AI Cisco has deployed, she sees if the workflows have changed, and if the renewals team demonstrated greater capacity since the AI assistant for renewals was deployed. That’s a much more telling metric for measuring success.
To truly get users to engage with AI, you must understand that agents are a continuous work in progress and they make users an integral part of the feedback loop as you iterate.
“The difference between an agent and a piece of software is with software, you build, test, deploy it, and it’s 100% feature complete,” Centoni says. “With an agent, it’s 60% of the way there in its first iteration in terms of usability, and the user is part of the tech stack that gives us feedback.”
A good channel for that feedback is essential, she adds, because you can only make an agent usable at scale when you get that feedback and continue to iterate.
For example, Cisco’s internal support agent runs seamlessly in the support workflow. As soon as a customer opens a case, the agent pulls in all of the customer’s relevant history and context so a support worker instantly has it at their fingertips, rather than spending the first 30 minutes with a customer asking for details of what went wrong.
“That’s a lot of time and frustration saved up front,” Centoni says. “It also provides the next best action. Now a support team member can help the customer and recommend things based on what is known about their environment.”
