• Network automation is high on the wish list of the telco community
• The first part of this feature looked at the drivers behind the trend and how a TM Forum evaluation methodology is helping the telcos with their automation plans
• But having a plan is one thing – overcoming the many challenges associated with network automation is quite another 
• “The long-term cost of AI-based automation is still a bit of an unknown,” notes TM Forum executive

As we established in the first part of this article on network automation, telcos around the world are starting to make efficiency and productivity gains from their early network automation efforts and many are using the TM Forum’s autonomous networks evaluation methodology to determine where they are with their plans and the next steps they need to consider – see The telco network automation journey – part 1.

But as Andy Tiller, EVP of member products and services at the TM Forum, told TelecomTV, there are many roadblocks standing in the way of the telcos’ network automation plans: He highlighted “the level of trust you put in the machines” as the most obvious challenge.

He noted that, based on the Forum’s evaluation methodology scorecard (see part 1 of the feature for more details), a level 4 autonomous network requires an operator to implement AI-based automation, which “effectively means rules generation.” Currently, explained Tiller, it is possible to have very sophisticated, rules-based automation where “somebody who really deeply understands the network” develops a set of rules that result in certain policies being implemented in given scenarios.

“That can be coded up into a sophisticated automation system, but it’s all been worked out by a person,” he pointed out. The next step, he noted, is when the AI generates new rules that run mission-critical networks without human intervention: That’s the level 5 ‘fully autonomous network’ with ‘closed-loop automation capabilities across multiple services, multiple domains (including partners’ domains) and the entire lifecycle via cognitive self-adaptation’.

Telcos are not at that stage yet – the telco community as a whole has an average autonomous network level of around 2.5 based on the Forum scale that runs from 0 to 5 – but that ‘fully autonomous network’ is the ultimate goal.

However, Tiller noted that the industry is “not yet at the stage where operators are happy to just let the machines control… Full AI-based automation requires a level of trust that I think the industry is still building,” he stated.

The knowledge gap

Another significant network automation hurdle is the knowledge gap related to the implementation of AI and how it is governed, managed and used correctly. 

Tiller noted that one way for the telco industry to develop this set of skills is through partnerships, particularly with hyperscalers. There is significant intent to address the knowledge gap, but “getting those skills in-house is definitely a challenge, because they are rare skills at the moment and they are expensive,” he explained.

Finally, Tiller pinpointed the difficulty of developing business cases linked to achieving higher levels of automation, such as level 4, which a number of major telcos, including Thailand’s AIS, China Mobile, MTN Group, Orange and Telefónica, aim to attain by 2025. “It’s a useful way to set a target, but some of them haven’t yet worked out why they want to be level four, and what the cost of being level four is. So, there may be no business case to get to level four in some areas. Level two may be absolutely fine because it’s not costing very much and it’s working,” suggested the TM Forum executive.

The cost of automation, however, and especially AI-based automation, may be prohibitive for some operators, and a return on investment is going to be very hard to quantify and achieve. “Really understanding where it makes sense to get to level four… that is something that is maturing in the industry. We are seeing operators take a sort of step-by-step approach,” he noted.

Tiller pointed out that some of the Chinese operators are “quite far ahead with this” and have reached level four – at least in their own self assessments for a number of scenarios that they’ve prioritised and for which they have identified a business case.

“They’ve implemented the automation – which obviously costs money to go from where you are to where you want to be – and they’re looking at the return on investment. But some of those costs are not yet fully understood, fully realised, particularly the cost of running AI-based automation. We’re still in an experimental phase, so the long-term cost of AI-based automation is still a bit of an unknown,” explained Tiller.

And while the Chinese telcos have a lead in the level of AI implementation at the moment, “the rest of the world is really watching and looking to catch up very quickly,” he concluded.

- Yanitsa Boyadzhieva, Deputy Editor, TelecomTV