From Noise Cancellation To AI Pods: Cisco’s Hidden AI Stack | Keith Griffin, Cisco

What does it take to build AI that cannot simply go offline when
the internet has a bad day? How do you bring large language
models, agents and deepfake detection into a 40 year old
networking company without breaking trust, security and
compliance? And why might the biggest career risk today be
ignoring AI rather than embracing it?


You can also watch this episode as a video podcast on my YouTube
channel.



In this episode of AI to the DNA, the podcast for everyone who
wants to really understand and apply AI, I sit down with Keith
Griffin. Keith is a Cisco Fellow, one of the highest technical
roles at Cisco, and at the same time VP for Collaboration,
responsible for products like Webex. He has been shaping Cisco's
AI strategy since 2016, co authored the company wide responsible
AI framework and works across engineering, M&A and research
to decide where and how Cisco bets on AI.


We talk about what really changes when you move from cool AI
demos to mission critical enterprise software. Keith explains how
Cisco thinks about outages, data residency and failover, why a
lot of speech plus LLM experiments fail in practice, and how you
design collaboration systems so that AI features run as close to
the problem as possible, from GPUs in meeting room devices to AI
ready data centers on premises. We look at deepfake risks in
video meetings, live deepfake detection, brand and safety
guardrails and why hope is not a strategy when you wire LLMs and
agents into real products.


From there we go into the workplace itself. Keith shares why
speech is coming back as a serious interface for work, how AI
note taking and agentic follow ups can fix very human weaknesses
like bad follow through, and what an "AI native" office might
look like when devices, network and agents all work together. We
discuss AI governance as a team sport between product, IT, legal
and procurement, and why the key soft skill for almost every role
will be asking better questions and iterating with AI.


I'm Christoph Magnussen and at AI to the DNA, I'm not just asking
how to use AI. I’m asking why it works the way it does so you can
gain insights on the nature of this technology. This podcast is
for everyone who wants to move beyond tool-tourism and master the
real craft of AI: Learning to think, work, and build side by side
with intelligent systems!


(00:00) - Hope Is Not a Strategy: Why AI Will Accelerate Work

(01:02) - Introducing Keith Griffin: Cisco Fellow & AI
Strategy Leader

(04:55) - Miscalculations, Timing, and 18 Years of Tech
Change

(06:48) - Enterprise vs. Consumer AI: Outages, Resilience,
and Trust

(10:25) - Cisco’s First AI Breakthrough: From Noise Detection
to Responsible AI

(12:36) - Deepfakes, Guardrails, and Why Security Becomes an
Arms Race

(17:48) - Building Agents for Real: Trust, UX, and Why
Single-Agent Comes First

(21:58) - Speech, Hardware, and Edge AI: Why Not Everything
Belongs in the Cloud

(28:19) - AI-Native Workplaces: Automation, Augmentation, and
New Work Patterns

(33:11) - Productivity in Practice: From Weeks to Minutes

(35:58) - The Biggest Mistakes Enterprises Make When Adopting
AI

(38:22) - Scaling AI Governance: How Organizations Keep Pace

(40:18) - Explaining AI to Kids: Schools, Skills, and Future
Careers

(43:58) - System Prompts, Agents, and New Attack Surfaces

(47:23) - Where Next? Why the Most Important Question Keeps
Changing