Op-Ed: Autonomous networks are coming sooner than you think. Much sooner.

  • Conventional wisdom in the comms industry held that human beings would be involved in approving network moves adds and changes for the foreseeable future

  • Huge advances made in automation technology, especially in AI and predictive analytics, have eroded that position

  • Any resistance to full automation will dissolve completely once first movers start to gain benefits in cost, performance, sustainability and reliability

In technology terms, it may not be up there with decoding DNA or splitting the atom, but the day when communications networks run themselves autonomously for the first time, without human involvement, is approaching. Based on the pace of development in AI and predictive analytics, we will cross this particular Rubicon much sooner than anyone anticipated.

“We are certainly moving towards fully autonomous networks,” said Erik Ekudden, SVP, CTO and head of Group Function Technology at Ericsson, in a conversation with me last week. “Some of the leading operators have the ambition to [implement] Level 4 autonomy by 2025.”

As a journalist, I like to refer to this as “news.”

Until very recently, conventional wisdom in the comms industry — especially the telecom bit of it — held that human beings would be involved in approving network moves adds and changes for the foreseeable future. Huge advances made recently in automation technology, especially in the areas of artificial intelligence (AI) and predictive analytics, have eroded that position.

Resistance is futile

The resistance will dissolve completely once first movers start to gain the radical benefits that full autonomous mode (a.k.a. Level 4 Network Autonomy) can deliver in cost, performance, sustainability and reliability. To stay competitive, their peers will have no choice but to follow suit.

To glean insight into what’s happening in the comms space, I compare it to a similar dynamic transforming military technology.

The U.S. Department of Defense’s formal position on AI is currently that a human operator will always be involved in decisions where humans are involved. But consider that the next wave of military conflicts will likely be fought with drone swarms — the only way to control them, or defend against them, will be fully autonomous AI. Once China, Russia, Iran or North Korea obtains this capability, the chance that the U.S. won’t follow suit is zero.

But in this instance, the U.S. war machine will likely lag the commercial sector — just like the U.S. telcos will likely lag behind more technologically advanced global operators.

First-mover advantage

All of the network automation vendors I have spoken to concur that it will be enterprises and verticals, such as manufacturing, transportation and energy, that will be the first to implement a full autonomous mode, rather than telcos, which are famously conservative about new technology.

When will A-Day happen? When I first predicted the advent of fully autonomous networks in 2017, the answer from most people was “never.” By last year, the window had narrowed to “a decade.”

I’m not going out on a limb by saying that the first implementations will be this year — yes, 2024.

Enterprises and service providers are already just one click (click!) away from full autonomous mode.

“The first stage of their journey is getting rid of performance issues on the existing network, stage two is optimizing the network, and the third stage is full automation,” said Mike Hicks, principal solutions analyst, ThousandEyes, which was acquired by Cisco in 2020.

ThousandEyes’ WAN Insights product uses AI and predictive analytics to continuously analyze performance data and apply statistical models to forecast conditions and recommend actions to ensure application performance. The product, which works on enterprise and communications service provider (CSP) networks, changes the mode of network and application management from reactive (Hey, what just happened?) to proactive/preemptive (Houston, we’re going to have a problem).

“We can automate now, but it’s still the customers’ preference to be asked first if they want to apply the recommendation,” says Hicks. “For now, they still want to see the workings, and get the verification.”

Customers approve actions by clicking on an icon on the WAN Insights management console. (Disappointingly, the icon is currently a sad grey-green color. I’ve asked Cisco to add a bright red “fully autonomous” icon to their dev plan. The company said it will think about it).

Trust issues

Autonomous network products have one last hurdle to overcome before customers push that A-Day button and it isn’t technological, it’s trust — both in the products themselves, but even more so in the vendors that sell them.

“[Autonomous capabilities] are not going to take over all these functions overnight. It will be more gradual because you need to build the [customers’] confidence that it actually does what it's intended to do, and that you can override it when you need to,” said Ericsson’s Ekudden.

Automation is central to the Ericsson Open Programmable Cloud Ecosystem (OPCE) strategy, and will also be a critical enabler of its enterprise and vertical industry strategy, sitting in its new enterprise division.

Trust is an area where companies like Ciena, Blue Planet, Cisco, Ericsson, Huawei, Juniper, Microsoft and Nokia are obviously heavily advantaged thanks to deep-seated relationships with the largest service providers and enterprise customers that in many instances go back decades. Conversely, customers are unlikely to buy from smaller vendors or startups.

“Getting bought by Cisco is the best thing that could have happened to us,” says Hicks, who is effusive about Cisco’s ongoing investment in ThousandEyes technology and the pivotal role Cisco has given the ThousandEyes’ portfolio in its new overall strategy.

If full autonomous mode still sounds far-fetched, it’s worth recalling that 15 years ago the idea that nine out of every 10 Fortune 500 companies would one day rely on an online retailer as an essential component of their digitalization strategy didn’t seem plausible either.

Jeff Bezos via Getty Images

Yet here we are, giving $24 billion a year to Jeff Bezos (a.k.a. Stupid Hat, see photo, right).

Which only leaves one question: Who’s first?


Don't miss our upcoming Cloud-native 5G virtual event, which will feature speakers from AWS, Microsoft, and Google Cloud. Register for your free pass here.