Hypergrowth AI Companies Need A New Kind Of Network For Employees

AI companies are on fire, clocking some of the fastest revenue growth in the industry. From new semiconductor startups to companies engaged in frontier AI models to companies fueling enterprise adoption of AI-enabled workflows, growth has been exponential. Startup and private company valuations have also grown, with many having raised record amounts of capital. While hiring and expansions have increased, most companies are choosing to work with a lean IT staff and prefer a high degree of automation. This has placed renewed emphasis on infrastructure agility and security.

Network planning used to assume a slower world. You looked at headcount and a few years of traffic history, projected the curve forward, and bought ahead of it with margin to spare. That works if you can see the curve coming and have people free to adjust as you go. AI-driven hypergrowth gives you neither.

Glean, the enterprise AI company, hit this early, mostly because it grew fast enough to feel it first. Glean is on a mission to connect knowledge, systems, and context, so AI can actually work. In a way, they are the "AI operating system" for their customers. Their speed, efficiency, and security have a direct bearing on several other customers.

Glean is a poster child for hypergrowth companies. What they expected from their wired and wireless networks and their expectations of network security are a fair proxy for several other companies that are experiencing such scale and tearaway growth.
 

Growth Is Surging While IT Teams Remain Lean

Glean's early network was run mostly by hand by a small IT team, and it was reactive and heavily patched. The team spent hours reproducing issues, troubleshooting individual locations, sending out workaround notices, and handling escalations. Bad Wi-Fi cut into meetings and generated a steady stream of tickets. The IT team was already stretched and adding more members to the team would not necessarily solve the problem.

As Glean's IT leaders put it, an AI-forward company can't afford to let foundational infrastructure turn into a manual bottleneck. For a company whose product depends on moving fast, the network was slowing everyone down.

Planning earlier wouldn't have fixed it because demand on an AI company's network does not ramp in a linear fashion. Someone connects a new AI application or model and traffic jumps within weeks, with no request filed and no warning. Headcount used to be a rough proxy for network load. It isn't anymore since an AI deployment can add serious activity without adding a single employee.
 

Infrastructure That Adapts To The Business

What Glean's IT leaders wanted was a secure network that stayed reliable and agile without the team having to keep it that way. They wanted a solution that was predictive, secure, and autonomous, offloading them from the day-to-day management and operational complexity that often overloads internal teams. IT teams today are increasingly expected to be nimble and strategic and rely on solutions built that way.

The network had to be equally secure by design. Security could not be an afterthought or introduce additional complexity or rigidity. This required a different architectural approach since legacy networks have security bolted on, which can introduce additional management overhead and cost and impede velocity.

Sunil Agrawal, Glean's CISO, framed the goal this way: "Building an Enterprise AI platform means every part of the business has to move with urgency and intention. Our infrastructure has to be as intelligent and autonomous as the products we build. When it is, our team stops managing infrastructure and starts advancing the mission."

After Glean moved to an autonomous as-a-service model, the office with the worst Wi-Fi problems saw downlink speeds go from around 28 Mbps to close to 100 Mbps, and 5 GHz channel utilization drop from roughly 75% to about 50%. Tickets fell by more than 90%. Complaints about the internet stopped appearing in the office Slack channel.

The team also stopped bracing for what each problem would cost. Under the old setup, a bad enough issue raised the question of whether it would turn into an expensive troubleshooting job or a hardware replacement.

What the team does with its time has also changed. Instead of reproducing issues and handling escalations, Glean's IT group is working on attendance tracking, automation with its own AI agents, ISP planning for new offices, and conference room standards.
 

What This Means For IT Leadership

Glean isn't unusual; the same thing is showing up at other companies where growth is exponential. It takes time for the network to stabilize and for the security posture to be trustworthy. But any change has a tendency to disrupt these very quickly.

It's very clear that what happened in the data center and cloud environments is now happening in on-premises deployments. The network is connective tissue and needs to be simple, adaptable, and inherently secure while delivering a stellar and predictable experience for IT teams, employees, and guests. This is not a future state, but rather an urgent present-day requirement fulfilled by modern Network-as-a-Service (NaaS) solutions that are secure and autonomous.