In certain instances, we may earn revenue when a reader makes a purchase or completes a form through our pages or on a partner’s website. We adhere to a strict 
Can machines take on real responsibility while leaving us entirely in charge? After days of listening, walking the show floor, and recording conversations at Cisco Live, the answer is quickly becoming yes. The next phase of artificial intelligence (AI) is not just clever conversation, but real action, delivered by intelligent agents that remember the context, pick the most effective next steps, and complete the work.
This shift is bigger than most people realize. It changes how companies operate, how employees perceive their roles, and how customers evaluate a brand. Here is what stood out to us and why this is not hype but a genuine and urgent business change.
Key Takeaways
- Agentic AI handles tasks, not just talk. Outcomes matter more than chat.
- Trust grows when AI proves real value, not promises.
- Context-rich agents stop repetitive errors and wasted time.
- Intelligent automation reduces burnout while leaving humans to make judgment calls.
- Secure agents act safely, protecting systems in the same manner as human administrators do.
- Start small, show clear wins, and then expand agentic workflows with confidence.
What Agentic Actually Means
In simple terms, agentic AI refers to software that performs tasks beyond merely answering questions; it is capable of acting autonomously. It understands its current position in a process, makes its own small decisions, and acts toward a clear outcome. Cisco has built much of its customer experience strategy around this idea.
Carlos Pereira, Cisco Fellow and chief architect at Cisco, described it plainly during a conversation with Techopedia:
“Everything inside customer experience for us is a workflow. When GenAI began, it was primarily focused on language. But our work is all about workflows. Agentic AI bridges that. It means you can break things into parts, each with an agent focused on a clear task.”
This comment explains why some businesses feel disappointed when they implement a chatbot on a website and expect immediate results. A language tool alone cannot achieve outcomes. An agent can.
Why Adoption Is Surging
Cisco’s research indicates that 68% of customer support interactions will involve agentic AI within a few years. In private talks at Cisco Live, many people told me that the estimate might be too cautious.
#CiscoLive: Agentic-Led CX is set to redefine customer experiences forever. 68% of all customer interactions are expected to be handled by Agentic AI by 2028.@CiscoLive pic.twitter.com/Pkl5NXUpPT
— NetCom Learning (@NetComLearning) June 11, 2025
Craig Burnham, who runs product marketing for Cisco’s collaboration tools, explained why companies are moving faster now:
“We know customer expectations keep rising. People want answers immediately. Our AI agent helps them get what they want quicker without waiting. At the same time, it helps human agents handle calls without burning out.”
This is the win-win hidden behind so much contact center frustration. Let the machine answer common questions instantly, but always keep skilled people ready when things get complicated.
Hype only carries new technology so far. Trust decides whether people use it every day. Numerous reminders were voiced at Cisco Live that trust requires objective evidence.
Burnham told Techopedia how Cisco builds this into every new feature:
“We test with real customers first. They want proof that the AI works. If it does not deliver, they stop using it. So we listen closely before rolling anything out.”
Carlos Pereira shared a story that reveals a great deal about this balance. A large media customer built a competent editorial helper, but staff refused to use it until blind tests proved it could draft simple updates just as well as humans. Once they saw that, adoption jumped overnight.
The future of AI success will be built on the feedback loop of trust between human design & desired AI outcomes. #ciscolive pic.twitter.com/nh8qvdesVp
— Adam Kain (@adam_kain) November 12, 2024
Context Is Everything
At Cisco Live, it was clear that agentic AI needs more than raw horsepower. Context matters more than clever conversation. Without it, any agent quickly becomes a frustrating black box.
Pereira explained how Cisco handles this behind the scenes:
“In our support tools, the AI assistant knows the full history. When we embed it inside the product, that same knowledge stays with it. So, the security team does not care if they talk to the product AI or the support AI. The context follows.”
This makes real automation possible. A user never has to repeat themselves just because they switch channels. It saves hours and lowers mistakes that cost money and reputation.
Around a quarter of Cisco’s global support cases were attributed to simple configuration errors. That is a massive waste of time and energy. But Pereira painted a clear picture of how agentic systems stop this at the source. He said:
“We map the whole data model. We run validation before any configuration change goes live. If something breaks, the system catches it early. We are serious about moving from configuration chaos to configuration confidence.”
For any large organization, this transforms AI from a fancy experiment into an insurance policy that protects uptime and keeps engineers focused on more critical tasks.
Humans in the Loop
Some leaders still fear that granting more autonomy to AI will result in fewer jobs. This came up in nearly every conversation I had at Cisco Live. Both Carlos Pereira and Craig Burnham addressed it head-on. Burnham put it in clear terms:
“It is about removing boring tasks that burn people out. Nobody wants to answer the same question all day. The AI handles routine calls. Agents handle the tough ones that need judgment and empathy.”
Pereira added how this plays out in engineering and support teams:
“We are not cutting people. We free them from admin work and repetitive checks so they have more time for critical cases. The value of human problem-solving goes up, not down.”
This attitude transforms anxiety into trust, explaining why many teams now suggest new agentic use cases themselves rather than resisting them.
Security & Resilience
Another consistent theme across sessions was how agentic models will inevitably raise the stakes for security. An intelligent agent that can change settings or open access must be protected as carefully as a human admin. Jeetu Patel, president & chief product officer at Cisco, captured this point perfectly during a main keynote:
“Safety and security define this AI era. Every new agent expands capacity but also creates a new entry point for risk. So, we build security directly into the network.”
From real-time tracking to zero-trust checks on agent actions, the goal is clear. Innovative tools must be as trustworthy as they are powerful.
Secure networking brought to you by #AI @Cisco @ChuckRobbins #ciscolive pic.twitter.com/Rsv3jkeaTU
— Zeus Kerravala (@zkerravala) June 10, 2025
How to Begin Without Overcommitting
Leaders I spoke to at Cisco Live offered simple, practical advice for companies wondering how to get started. Do not chase hype. Pick a clear problem where agents can help. Measure the benefit. Then grow carefully. Pereira summarized it like this:
“People want perfect from day one. That is unrealistic. Pick a well-known process, automate it with a clear agent, watch the results, and then expand once you trust it.”
Craig Burnham echoed the same idea for contact centers: start with high-volume routine questions. Once confidence is established, introduce more advanced actions and multi-agent workflows.
Another insight from the event is how Cisco stitches all these tools together. Instead of separate dashboards for each agent or feature, Control Hub consolidates them into a single view. Teams see precisely how much AI is used, where it adds value, and what needs fine-tuning.
Burnham described this in our chat:
“It is about showing customers clear adoption and return. Everything lives in one place. They do not have to hunt for reports or manage separate tools.”
This unified approach keeps complexity down, builds trust, and helps leaders scale agentic ideas without hidden surprises.
What’s your biggest takeaway from #CiscoLive this year? pic.twitter.com/6KBqljhjb1
— CBT Nuggets (@CBTNuggets) June 12, 2025
The Bottom Line
Leaving Cisco Live, I kept returning to one question. When every vendor says they have AI, how will customers decide who to trust? For me, the answer is companies that build clear guardrails, share proof of value, and always keep people close to the loop.
Agentic AI is not a gadget. It is an upgrade for how work gets done. The winners will be the ones who treat it as a teammate with clear rules, not just an add-on script.
