“We’ve all seen AI explode in the last two or three years,” said Cisco CEO Chuck Robbins to a packed house of AI visionaries, global business leaders, and technology innovators last week.

But Robbins was quick to add a sobering stat from Cisco’s AI Readiness Index: Only 13 percent of enterprise customers feel they have a firm grasp on their artificial-intelligence strategy — while nearly 90 percent feel pressured to figure it out in the next 18 months.

The high stakes and dazzling pace of AI innovation were key themes at Cisco’s AI Summit in Palo Alto, Ca. And with a world-class collection of experts, the event shed some much-needed clarity on a wide range of AI-related developments. That included everything from industry applications and forthcoming innovations to ethics, governance, and societal implications. Not to mention, a groundbreaking product release from Cisco. AI Defense tackles the complex security issues that come with the development and use of AI applications.

Cisco EVP and chief product officer Jeetu Patel served as master of ceremonies throughout the day. And he began his keynote with a quick summary of Cisco’s comprehensive innovation and support in the AI space. This includes the critical infrastructure to handle the massive demands of AI — with high bandwidth, low latency, and reduced energy consumption — as well as the services and support to implement and drive successful AI transformation.

Because getting AI right — and meeting the challenges of security, energy demands, talent shortages, and more — can reap tremendous benefits.

“I see all things abundance and growth and innovation from AI,” said Aaron Levie, CEO of Box. “We're in a period where just about every day there's some new breakthrough technology and AI underpins all of it. But imagine just 20 years from now, what it could mean for quality of living, energy production, and self-driving technologies. And what we’re going to be able to do in life sciences, biology, physics, and so much more, using AI and data.”

Safety and security, but not at the expense of speed

After his keynote, Patel dove into security. While he stressed that the AI Summit would highlight a wide range of thought leadership in the AI space, the event was a perfect opportunity to introduce the features of Cisco AI Defense.

As it regards AI, he said, “it’s astonishing to see that things that felt like magic just a couple of years ago now feel commonplace.”

But for many companies, it’s still a constant struggle to catch up. Employees are increasingly using external AI, while their own developers rush to create new proprietary AI apps. All as security fears mount.

AI Defense secures the development of AI applications while simultaneously protecting from external AI vulnerabilities — and does it all at scale and at lightning speed.

“We can't afford to sacrifice safety and security for speed,” Patel stressed, “but you also can't afford to sacrifice speed for safety and security. You’ve got to have both.”

Cisco AI Defense uses algorithmic red team testing and robust runtime guardrails to protect AI applications throughout development and deployment. While it might take a human weeks to perform these vulnerability tests, AI Defense works in seconds. Moreover, AI Defense protects against employee use of unsanctioned and potentially unsafe external AI tools.

“We enable organizations to ensure that their employees can access applications that they use for AI, for their productivity for the day-to-day work safely and securely,” Patel further explained. “That's the first problem that we help solve. And then the second problem is when you're building an AI application, how do we make sure that that AI application has the right levels of safety and security factored into the model.”

Best of all, these powerful capabilities are built into the Cisco Security Cloud, are simple to use, and they utilize Cisco’s proprietary machine-learning models and threat intelligence data from Cisco Talos. That means AI Defense can evolve and adapt to ever-changing safety and security concerns.

After demonstrating its intuitive dashboard and powerful features, DJ Sampath, Cisco’s vice president of product, AI software, and platform, summed up what makes AI Defense special.

“Being able to do model validation at scale is a huge differentiator for us,” he said. “And being able to do the enforcement at scale now, to be able to enforce this wherever customers are deploying the models, wherever customers are using these models, that's really a game changer.”

Optimism, tinged with caution

A key theme was AI’s potential to augment human efforts, not replace them. A company called Augment Code is doing just this for software engineers. And Scott Dietzen, Augment Code’s CEO, shared some of the challenges these engineers currently face, including maintenance testing, documentation, disaster recovery, fall tolerance, security requirement, and more.

“We're trying to augment software engineers to make their lives easier,” explained Dietzen, “to deliver a 5x or even a 10x more productive software engineer, by removing some of the drudgery and pain associated from advancing a really large, complex piece of software.”

On a similar note, David Solomon, CEO of Goldman Sachs, told Chuck Robbins of his early days in banking, when research meant combing through stock tables in the Wall Street Journal at the local library.

“Today I can say it into a supercomputer, my phone, and I have it instantaneously,” he said. “It doesn't mean we have less smart people working at Goldman Sachs. It just means the productivity level of what they're doing, the focus of what they're doing continues to expand. And that grows the business.”

While he’s “super optimistic” about AI, he sees a level of trepidation among his peers.

“I find a sense of enthusiasm with AI,” Solomon added, “but with caution. Because you want to make sure you get it more right than wrong. It can be a safety issue; it can be a cost issue. But the cost curve is going to accelerate down, I think pretty quickly over a period of time.”

Adding to the optimism was Arianna Huffington, author, Huffington Post founder, and CEO of Thrive Global. She has been extremely active in adopting AI to promote healthy habits and work/life balance. And while she’s outspoken about the online negativity that appeals to people’s worst traits, in a conversation with Fran Katsoudas, Cisco’s EVP and chief people, policy, and purpose officer, she expressed her belief that AI can help us do better.

“The new forms of AI using hyper-personalization,” Huffington predicted, “can become a defense against the earlier forms of AI that tended to tap into the worst in us, our biases, our rage, our anger.”

“So, what about companies appealing to what's best in us,” Huffington continued, “appealing to the better angels of our nature, appealing to what makes us want to give and to be healthy and to collaborate and to empathize? I think we can demonstrate that you can make just as much if not more money doing that.”

The best way to govern AI? With AI.

While some at the AI Summit expressed their antipathy to AI regulations, others saw it as an essential enabler for success and trust.

Dev Stahlkopf, Cisco’s chief legal officer, welcomed to the stage Navrina Singh, the founder and CEO of Credo AI. Her company has developed an AI governance platform that empowers organizations to deliver and embed artificial intelligence responsibly. In short, it uses AI to manage, measure, and monitor AI. This simplifies and streamlines governance, which Singh called a highly misunderstood, but essential process for making AI successful in business.

“One of the things that we are seeing evolving in this ecosystem is how do we actually first and foremost make sure that the governance tooling is able to be that single pane of governance across all your AI ops tooling,” she explained. “And that's where integration becomes really critical.”

Singh expects AI-generated governance to be widely adopted.

“In the next three years,” she said, “AI governance is going to become your strategic advantage. It is not going to be like, ‘let's do it later’; it's going to be, ‘how much are you investing in it’?”

What’s to come in AI

A number of experts spoke of the rise of “agentic AI,” in which today’s AI agents continue to evolve, becoming, in effect, extra employees, working and conversing with humans to solve problems, set agendas, and solve problems, increasingly with their own proactive independence.

“We very much believe in this augmentation model,” said Arvind Jain, CEO of Glean, “a world where we are thinking about AI as a tool as opposed to AI that replaces us. Ten years from now, I believe that 90 percent of the work where we spend our time today, we won't be spending anymore. There's going to be for every individual this amazing team of AI agents around you that serve as your assistants, your coworkers, coaching and doing work for you.”

And what about the whip-smart, humanoid robots of sci-fi fame? Fei-Fei Li, the legendary Stanford University scientist and founder of World Labs, is working on 3D modeling that could anticipate a next step in human/machine interactions.

“We believe that the ability to generate and interact with 3D worlds, real or virtual, is fundamental for intelligent agents, whether it's people or robots or virtual agents,” she explained. “It unlocks a lot of applications and businesses, because it's just so natural for an intelligent agent to work in that kind of space. So, we want to do the next generation foundation model for that. Because visual intelligence is a cornerstone of intelligence.”

Li sees challenges in creating these 3D visual models. For example, they lack the deep data that led to fast advances in large language models. But as Li stressed — and the entire event confirmed — progress in AI shows no signs of slowing.

“Things are happening now in AI that I never anticipated,” she concluded. “And the younger generation is going to make so much more happen. It’s just an exciting time to be in tech.”