Image by Markus Winkler on Unsplash
Most companies right now boast about “owning their data” like it’s their secret sauce for AI success. But according to Cisco, that’s just talk — and they’re not shy about calling it out.
I sat down (well, virtually) with insights from Jeetu Patel, Cisco’s EVP and Chief Product Officer, and DJ Sampath, who heads up AI Software and Platform. Cisco’s take? Unless your AI models are learning from machine data — the logs, metrics, and telemetry that systems pump out constantly — you’re missing more than half the story.
The data you’re ignoring is growing faster than everything else
Right now, a whopping 55% of data growth in enterprises comes from machine data. Yet, most AI models are still trained almost entirely on publicly available, human-written content.
You know, internet text. Articles. Social posts. The stuff we’ve already squeezed dry.
So where’s the next frontier for smarter, more useful AI? It’s hidden deep inside enterprise systems, generated by machines, and that’s what Cisco wants to unlock.
“We’re done with public data,” Patel explained. “Where else do you go next? It’s all locked up inside enterprises.”
If AI engines are cars, machine data is high-octane fuel — and businesses are sitting on treasure without the pipelines to tap it.
The line between model and product has vanished
Patel says the classic divide between “model companies” (think OpenAI) and “product companies” is fading fast. In the very near future, every successful company will be both.
“It’s a loop,” Patel said. Products shape models, and models shape products. Companies that only build UIs on top of third-party AI won’t last. They’ll get outpaced by those who build their own models — trained on their own internal machine data — and adapt them as product demands shift.
Sampath took that further: “Agents are next. But the strength of an agent depends on the model behind it. The model is your moat.”
Why Cisco thinks hardware isn’t dead — it’s your secret AI weapon
Image by Markus Winkler on Unsplash
Now here’s an unexpected take. In an era where software supposedly eats the world, Cisco is doubling down on hardware.
Why? Because hardware generates the very machine data that’s missing from most models. And Cisco knows how to collect, organize, and use it.
Take infrastructure maintenance. According to Patel, Cisco’s models can now tell when a one-degree shift in hardware temperature might signal a system failure in three days. That means you can preempt outages, reroute traffic, and stay online. Try pulling that off with a traditional AI model trained only on public language data.
That kind of predictive power isn’t just useful — it’s a strategy-level advantage, especially for manufacturing and asset-heavy industries.
Open source, but make it secure
Security is where Cisco’s AI philosophy really shows up. Instead of hoarding their best models, they open-sourced them. Foundation-Sec-8B — a powerful security model — has already seen 200,000 downloads on Hugging Face.
Why give it away?
“Attackers already have access to open source AI,” Sampath said. “So we’re making sure defenders have strong AI models too.”
They’ve also integrated ClamAV, a widely-used antivirus scanner, to check every model on Hugging Face for malware. That’s next-level supply chain security — and yeah, it matters when your future workforce might involve 100 AI agents per person.
Customers are split — and that’s normal
Cisco just wrapped up its Cisco Live event with a suite of new tools, and customer reactions fall into three buckets:
- The enthusiastic adopters — “We’ve been waiting for this. Finally!”
- The cautious testers — “Looks interesting, let’s see how it works.”
- And the old-school skeptics — the “let’s verify every single detail” crowd.
Interestingly, the third group has been shrinking, says Patel. As products ship faster and actually deliver, trust levels have climbed.
Cisco’s biggest challenge now? Keeping up with themselves.
“We’ve got so much innovation in the pipeline that our main job is just keeping customers updated,” Patel said.
A quiet strategy behind the shift
This isn’t about a wholesale rip-and-replace. Cisco’s approach isn’t to throw out old systems, but to meet customers where they are and walk with them — firewall upgrades today, multicloud security tomorrow, AI-powered smart switches when they’re ready.
Behind all this is a platform they call Security Cloud Control, layered with tools like AI Canvas and Hypershield. The goal? A modular, manageable rollout of next-gen AI capabilities without overwhelming IT teams.
What’s next: turning partners into AI accelerators
Looking ahead, Cisco is gearing up for its November Partner Summit in San Diego. They’re calling on global partners to help bring this AI approach to more organizations — especially those who can’t figure it all out on their own.
As Patel puts it: “The reseller engine needs sustained acceleration.” He’s placing long-term bets on partner ecosystems that can translate Cisco’s tools into real-world, customized AI deployments.
—
So here’s the thing: Cisco isn’t just thinking about how to survive in an AI future. They’re betting the company on something deeper — that AI driven by machine data, hardware intelligence, and open-source security will give businesses an actual edge, not just buzzwords.
And if they’re right, the successful companies of tomorrow won’t just build products. They’ll build models — trained, tuned, and fed by data that only they own.
—
Keywords: machine data, AI strategy, Cisco AI, enterprise data, Foundation-Sec-8B, open source security, hardware and AI, AI agents, enterprise AI, Security Cloud Control
