Contributing Editor Annie Turner rounds up the summer’s news headlines and notes how AI is moving beyond ML and positioned to unlock telcos’ data to speed automation

Heavy Reading has published a timely and, as far as we know, the first study into the future of network automation for Tier 2 and Tier 3 network operators. They are defined here as having revenues of less than $1 billion annually. The survey was commissioned by Juniper Networks, with responses from 217 Tier 2 and 3 execs around the globe – a far bigger sample than the average study.

For 30% of those surveyed, most operations are almost entirely manual, relying on command-line interfaces and other legacy software: closed loop and higher levels of automation are very rare, regardless of the size of the operator within the definition. Only 6% say they have fully autonomous, self-driving network automation.

The issue isn’t that these operators don’t understand the potential of automation for both customer-facing and internal, network-centric purposes, as the graph shows.

It’s just that it’s too difficult for a whole lot of reasons, not all of which have to do with technology. For example, the third biggest hurdle (named by 29% of respondents) is internal resistance due to the fear the automation will cut jobs and another is a lack of clarity about automation’s objectives. The the report looks at the role of SaaS, cloud and AI in network automation. Read on for how the elements of the industry are, via various routes, striving to address some of those obstacles.

Democratising AI?

VMware made a clutch of announcements at the opening of its VMware Explore event in Las Vegas in August. They included “new Private AI offerings to drive enterprise adoption of generative artificial intelligence,” and expanded its relationship the world’s biggest maker of silicon for AI, NVIDIA. It launched the VMware Private AI Foundation with NVIDIA. This, VMware said, puts the power of generative AI within the reach of any enterprise.

The chip-maker’s CEO and founder, Jensen Huang, echoed this in his presentation at the event: “Our expanded collaboration with VMware will offer hundreds of thousands of customers – across financial services, healthcare, manufacturing and more – the full-stack software and computing they need to unlock the potential of generative AI using custom applications built with their own data.”

Breaking data barriers

This interesting in the light of a the Heavy Reading survey and one carried carried out by Analysys Mason earlier this year of 84 telco execs from around the world. It was commissioned by Nokia and the main finding was that one of the biggest hurdles is integrating AI into heterogeneous networks is that access to high-quality data sets is hampered by legacy systems with proprietary interfaces. It also found that high costs mean some operators are not investing in AI platforms.

Enter the Global Telco AI Alliance. It was formed at the end of July by Deutsche Telekom, e& and Singtel to accelerate telcos’ AI transformation in telcos and develop new AI-powered business models. The memorandum of understanding between the parties includes the co-development of a Telco AI Platform.

Just a couple of weeks later, SKT again underlined its commitment to AI and to producing assets for use by other operators, announcing it had invested $100 million in AI firm Anthropic to build a large language model (LLM) customised for telcos. Anthropic had already raised $450 million from Google and Spark Capital in May.

VMware edges closer to Broadcom

Circling back, VMware seems closer to being acquired by Broadcom, having secured approvals for the proposed deals in many countries, most recently from the UK’s Competition Market Authority in August. The proposed deal was announced last summer but wasn’t greeted by much enthusiasm by Wall Street which didn’t see the logic of the transaction.

One school of thought is that chip-maker Broadcom plus VMware’s cloud and virtualisation software could bridge the gap between the data centres it powers and the applications that run in them. As Tom Nolle writes, “The data center used to be the focus, the resource that ran everything. Today, I’d argue that virtualization and the cloud have gotten us back to where our focus needed to be – the applications that support businesses.”

It’s worth noting that China has yet to pronounce on the proposed acquisition. There are suggestions that if it objects, the deal could be scuppered. Broadcom’s CEO, Hock Tan, has dismissed those concerns.

Finally, back in the here and now, BT is running a proof of concept (PoC) with French firm Infovista which specialises in network lifecycle automation. The aim is to reduce interruptions to voice services through faster identification of root causes, shorter mean times to resolution and greater customer satisfaction. Infovista said it generally expects to cut resolution times by more than two-thirds for network operators through an automated workflow.

In the PoC workflow detects issues, carries out analyses and generates alarms within BT’s trouble-ticketing system. The alarms embed data to streamline troubleshooting of issues in the network and services, and that are subscriber related. The tech should also reduce operational costs as there is no need for level 1 support or fast-tracking issues to the right level 2 teams, according to the vendor.

This is about shifting from manual to automated operations that are governed by key performance indicators and service level agreements to maintain service and business objectives.