AI demands network agility and better operational governance

Enterprises are grappling with massive change in the way applications are used and how information is exchanged. The network and security infrastructure of yesteryear does not cut it in the cloud-first or multi-cloud world. Organisations need to adopt next generation network and security technologies so they can move with more agility at a fraction of the cost of traditional legacy networks.

Tim Sullivan, the CEO and cofounder of Coevolve said we are in the midst of a revolution.

"The previous experience when people were starting to adopt cloud with a variety of different providers as well as their legacy applications created a whole new set of traffic patterns. Now AI is changing latency sensitivities and the amount of traffic on the network. We're seeing scores of agents added every week increasing the amount of traffic and changing flows."

Making AI work demands a new way of looking at infrastructure and security in Sullivan's view.

This new operating model starts by assessing the required internal skills and whether specialist provider support is needed. Organisations need to invest in change management and governance processes within a software-defined architecture. They need strategies to manage the potential emergence of shadow AI.

The rapid adoption of AI technologies, driven by compelling use cases, introduces significant risks. Organisations need tools that enable effective enforcement of responsible AI policies with robust monitoring mechanisms to ensure adherence to approved tools and prevent data leakage through unauthorised applications or plugins.

"Things are happening at machine speed now," said Sullivan. "And it's not just good corporate or government users with these tools. Bad actors have access, so people need to act faster and ensure they have the right governance in place."

Successfully managing increasingly complex infrastructure while countering a rising tide of threats puts observability front and centre. Organisations need greater ability to scan across highly distributed environments where they have a raft of different assets. Sullivan says there's a tendency, in the early adopter stage, to focus on the technology. But there's less focus on the less sexy part of how the new technology will be supported.

Research conducted by Coevolve put the spotlight on this issue. In 2025, just over a third of organisations were experimenting with AI. That has now jumped to over 70% of organisations. But, at the same time, the number that say they have appropriate governance in place has barely moved from 36% to 37%. In Sullivan's view, this indicates many organisations are not ready to scale their AI endeavours.

"AI comes with all sorts of promises of unlocking great levels of new productivity or new services. It really is quite a revolutionary time," he said. "But adoption has raced far ahead of any notion about end-to-end visibility and people understand what's responsible use versus naive or malicious use."

As organisations start to embrace AI fully, Sullivan says they make assumptions that may become their undoing.

The first is the expectation of ubiquitous, high quality, resilient connectivity when there are more demands being placed on networks. AI will not only drive increased traffic volume, but the destination, flows and latency demands are changing.

"There can be a brownout where there is packet loss on a particular part of a network. At sites that we're managing for our clients, the software measures all sorts of metrics ten times a second. If a connection is not working optimally, traffic can be directed through multiple paths to minimise disruption," Sullivan added.

Knowing what users are doing is a blind spot he said. Many users are not properly trained, monitored and governed. Sullivan said the only answer to this is to work at machine speed and adopt a zero-trust architecture to minimise the risk of an incident.

A key to ensuring networks are robust and agile is to adopt a software defined model. This enables performance improvements, operational ease and the ability to deploy and enforce policy changes quickly. And it drives down cost.

Doing this successfully means organising teams that are not bound by traditional operational silos.

"You've probably got a team that does the network infrastructure and a team that does the security and secops. If that technology is converging you need to reconsider your people management plan," Sullivan said.

For global companies, the temptation might be to look at one of the major global telcos. But that is changing Sullivan said, Increasingly, he is seeing large telcos home in on specific target markets and move away from trying to deliver global connectivity. That means a software defined infrastructure becomes increasingly critical so the services from different providers can be stitched together.

Enterprises face significant challenges adapting to the accelerating pace of change driven by AI proliferation and the demands it places on networks. Successful navigation of this environment demands a proactive approach centred on observability, robust governance, and a flexible operating model. Organisations must prioritise investment in skills development and strategic partnerships to mitigate risks associated with rapid technology adoption and maintain agility in a complex, increasingly dynamic landscape.