- Enterprise IT environments have grown into sprawling megacities, with the average company running apps across 19 locations, F5 CEO François Locoh-Donou said
- AI is making it worse: enterprises are managing an average of seven AI models in production
- One billion AI agents will be deployed by 2029, Locoh-Donou said, at the opening keynote of F5 AppWorld 2026 in Las Vegas
F5 APPWORLD 2026, LAS VEGAS — Enterprise IT environments have grown into sprawling megacities, and AI and digital sovereignty are about to send growth into overdrive, F5 President and CEO François Locoh-Donou told attendees at F5 AppWorld 2026 Wednesday.
“Enterprise IT environments are not unlike a city. Every infrastructure environment, every data center, every cloud region, every colocation facility, is like a neighborhood in that city,” he said in his opening keynote. Apps are like buildings. Traditional applications are monolithic structures. Microservices are like residential complexes. APIs are roads connecting buildings, crisscrossing neighborhoods.
“Your data: that is the power grid required in every thriving city,” Locoh-Donou said.
Like cities, IT environments grow over generations, often without planning. And like cities, the professionals tasked with managing the environments inherited the metropolis “whether you wanted it or not,” the F5 boss said.
As an illustration of IT environments’ complexity: F5’s latest State of Application Services research found that enterprises today operate applications across an average of 19 locations — typically five on-premises data centers, five colocation facilities and nine cloud regions. Ninety-four percent of enterprises operate in hybrid, multi-cloud environments.
Now, two forces are about to expand those cities at exponential speed: AI and digital sovereignty.
AI is already in production
Two years ago, the assumption was that most enterprises would consume AI by pointing APIs at cloud and model providers. That’s not what happened. Instead, 78% of enterprises are self-managing their AI inference workloads, Locoh-Donou said — largely for privacy reasons, since handing data to cloud providers means those providers can do essentially whatever they want with it.
That puts delivery and security for AI squarely in the hands of enterprise IT teams, he said.
And it’s not just one AI model to manage. Enterprises have already standardized on an average of seven generative AI models in production, choosing the best model for each task category — the same logic that drove the adoption of hybrid multi-cloud in the first place.
Agentic AI accelerates that complexity. An IDC study projects that by 2029, more than 1 billion AI agents will be actively deployed — over 40 times the number deployed today — taking 217 billion actions every day. Each of those actions needs to be delivered and secured.
A Gartner study cited by Locoh-Donou predicts that by the end of 2025, 40% of enterprise applications will integrate AI agents, up from just 5% a year ago. By 2028, those agents will intermediate more than 90% of business-to-business purchasing — $15 trillion in transactions flowing through automated, agent-to-agent exchanges.
And the cybersecurity landscape has gotten hotter recently, as the Iran war, quantum computing advances and agentic AI complicate the already difficult IT security landscape, F5 field CISO Chuck Herrin told Fierce in an interview earlier this week.
The API visibility problem
Every time an AI application calls a model or an AI agent calls a tool, an API does the work. APIs are multiplying fast, and many are invisible to security teams.
“You can’t secure the roads you can’t see,” Locoh-Donou said.
F5 is extending its API discovery and security capabilities from its distributed cloud platform to BIG-IP, closing what he called a massive blind spot. With 86% of enterprises deploying applications across clouds, edge, and on-premises environments, a SaaS-only option wasn’t enough. The company is also offering a fully air-gapped API discovery and security solution in early access.
F5 announced several new capabilities at the conference.
F5 Insight for ADSP uses natural language through popular LLMs, or an enterprise’s own compatible LLM, to give application owners plain-language reports on application performance, stress points and systems needing attention, without requiring deep BIG-IP expertise.
On the AI security front, F5 acquired Calypso AI several months ago, giving it both offensive and defensive AI security capabilities. AI Red Team agents simulate thousands of attack patterns against customer models daily, adding more than 10,000 new attack techniques to the library every month. AI Guardrails lets users build custom policies to address those threats in real time. F5 AI Remediate, which interprets AI Red Team findings, recommends the appropriate guardrails, and lets customers activate them, automating a process that previously required scarce AI security expertise. BIG-IP release 21, also announced at the show, adds native support for S3-compatible storage and MCP, two protocols Locoh-Donou called critical for the AI era.
All of these capabilities are being brought together under F5’s Application Delivery and Security Platform (ADSP), introduced at AppWorld last year.
For service providers, the most significant launches from F5 are Remediate, as well as Insight tools for visibility and analytics, and cloud and AI security tools, Richard Duncan, F5 senior director of solutions engineering for major service provider accounts, told Fierce. Automation is the common thread among all of these.
“Service providers are under a lot of pressure to do a lot more with less. They’re not hiring more people,” Duncan said.
Digital sovereignty adds new building codes
The second major force reshaping enterprise IT is digital sovereignty — regulations governing where data can live, how quickly organizations must recover from disruptions and which infrastructure providers they can rely on, Locoh-Donou said.
In Europe, DORA and GDPR are already law, affecting large enterprises across Europe and the Middle East. Many global companies are adopting the most stringent regulatory requirements as their global standard rather than fragmenting operations by country.
The U.S. has different requirements, but resilience mandates are pushing more hybrid multi-cloud architectures here too.
The net result, Locoh-Donou said, is that the already complex enterprise IT megacity is getting new building codes on top of everything else.
His prescription: a single, interconnected management platform that can see every building in every neighborhood, secure every road, and enforce building codes across the entire city. That’s what F5 is building with ADSP, Locoh-Donou said.
“Cities are messy, they’re complex,” Shawn Wormke, F5 SVP of product marketing, said in his keynote, which followed Locoh-Donou. Wormke described how he grew up in Gillette, Wyoming, a relatively small town, and moved to Chicago as an adult and came to love cities. “They are always evolving, yet somehow millions of people move through them every day, safely, reliably, predictably,” he said.