Enterprises need to plan for AI agent hosting and network connectivity with flexibility in mind.
Enterprises aren’t totally sold on AI, but they’re increasingly buying into AI agents. Not the cloud-hosted models we hear so much about, but smaller, distributed models that fit into IT as it has been used by enterprises for decades. Given this, you surely wonder how it’s going. Are agents paying back? Yes. How do they impact hosting, networking, operations? That’s complicated. Right now, of 394 enterprises who have offered comments on their experience, 47 have fairly extensive deployments to talk about, and it’s this group that can give us the best answers to those questions.
Let’s start with hosting, and the key point here is that resource pools to host AI agents are not the answer. There are multiple reasons for this, but the main one is that the hosting location for AI agents has to balance being close to the databases used, the applications the agent is part of, and the users. Which is more critical depends on the specifics of all these issues, and so there’s no common solution. That makes it really hard to cluster AI agents in a single place. In addition, the best agent technology may well differ across AI agent applications, so creating a single, giant multi-server AI model may not be the solution.
Enterprises are also finding that the database use of an AI agent often evolves, both because the mission changes for an agent as much as for any software component, and because they find new information resources the agent needs to do its job properly. Because an AI agent can do things traditional software components can’t, it needs to “know” more things, which means using broader sources of information resources than expected. Having different data sources means potentially different optimum agent hosting locations.
That creates the primary network issue for AI agents, which is dealing with implicit and creeping data. There’s a singular important difference between an AI agent component and an ordinary software component. Software is explicit in its use of data. The programming includes data identification. AI is implicit in its data use; the model was trained on data, and there may well be some API linkage to databases that aren’t obvious to the user of the model. It’s also often true that when an agentic component is used, it’s determined that additional data resources are needed. Are all these resources in the same place? Probably not.
The enterprises with the most experience with AI agents say it would be smart to expect some data center network upgrades to link agents to databases, and if the agents are distributed away from the data center, it may be necessary to improve the agent sites’ connection to the corporate VPN. As agents evolve into real-time applications, this requires they also be proximate to the real-time system they support (a factory or warehouse), so the data center, the users, and any real-time process pieces all pull at the source of hosting to optimize latency. Obviously, they can’t all be moved into one place, so the network has to make a broad and efficient set of connections. That efficiency demands QoS guarantees on latency as well as on availability.
It’s in the area of availability, with a secondary focus on QoS attributes like latency, that the most agent-experienced enterprises see potential new service opportunities. Right now, these tend to exist within a fairly small circle—a plant, a campus, perhaps a city or town—but over time, key enterprises say that their new-service interest could span a metro area. They point out that the real-time edge applications tend to exist where multiple facilities are close enough to be interdependent but too separated for wires, Wi-Fi, or even private 5G to be useful.
There’s a nice bridge to our last topic, operations, and here the key point is you’ll need more than an ounce of prevention, because a cure may not be possible. It’s critical to plan out AI agent hosting and network connectivity with flexibility in mind, and to consider any likely paths of agent feature evolution when working through hosting and connectivity plans. What an AI agent needs for data, and what workers or processes will depend on it – these requirements are much more easily accommodated before servers are installed and services are purchased. If improper design of AI hosting, and inadequate network preparation, result from poor planning, the issues they cause can be difficult to detect and impossible to remedy without considerable expense.
Planning is the key to AI agent success. AI agents may be used like software components, but the fact that they’re supposed to learn and know, rather than just to do, means they evolve more, and more quickly, than traditional software. If this kind of spontaneous expansion isn’t planned for, the result can be poor quality of experience and exploding costs.
