- Plenty of global CSPs are testing AI, but few have disclosed specific metrics to show how far along they are to autonomous networks
- Implementing AI in NetOps is a key challenge, because one mistake could impact a large number of subscribers
- Providers across the board right now are mostly using AI to improve customer service
We hear every day about how artificial intelligence (AI) can be used across different industries. In the comms industry, there is no shortage of news about communications service providers (CSPs) moving headlong into AI. However, when it comes to internet service providers (ISPs), they’ve still got a ways to go — and they have several approaches to take, according to Jason Presement, business advisor of market expansion at Canadian Fiber Optics.
Just like CSPs, ISPs can use AI for network optimization, customer service, operational efficiency, predictive maintenance, fraud detection, revenue growth and sustainability. Essentially, the world’s their oyster. In fact, CSPs have been predicting service outages and maintenance for a while.
"It just hasn’t been called AI," Presement said. “It’s just an automation. It’s a collection of data and it’s been predictive modeling…there’s more of that happening now," he said at an FBA webinar this week, referring to “identifying patterns and trends, anticipating failures and disruptions before they happen.”
AT&T, Bell, Rogers and Telus are some notable CSPs tapping into AI to improve customer service, Presement noted.
However, few CSPs are “revealing concrete metrics” to show how far along they are in their public goals toward autonomous networks, said AvidThink analyst Roy Chua, but some stand out, particularly outside North America. China Mobile for instance is among the operators with the “most aggressive” AI targets, Chua told Fierce. It’s aiming for level 4 autonomy — where systems can make decisions without involving humans — across multiple telco domains by 2025, he added.
China Telecom, Orange and Telecom Argentina are some other AI trailblazers, he added, along with “less well-known players” like Malaysia's Digital Nasional Berhad and Indonesia's Telkomsel.
ISPs and the AI challenge
Not all CSPs have “sufficiently modernized” their data or operational systems, Chua said.
While Chua didn't have specific details about ISPs, he did say that, in general, accessing data is one of the most important aspects for benefitting from AI.
“Key to using AI/ML or generative AI to help with these tasks is the ability to gain access to data and metadata efficiently,” said Chua explained. “Some service providers will lack in-house expertise around AI/ML/GenAI to effectively deploy these technologies and will need to lean on external partners for assistance.”
For sure, dealing with disparate systems, “different data sources and people collecting and reporting and talking about different data in different ways,” is one challenge for an ISP, Presement said.
It’s existed since the old network management days and “it’s starting to look like that in the AI realm as well," he added.
"Failing to set the foundation with AI-ready data is one of the biggest challenges that ISPs and other service providers face," said Susan Welsh de Grimaldo, senior director analyst at Gartner. She noted developing AI projects that deliver desired results "depends on the data inputs and clearly aligning AI strategies with business goals."
One AI use case that’s often brought up for ISPs is for troubleshooting and suggesting root causes for outages. But Chua doesn’t expect “fully autonomous closed-loop systems that can automatically remediate outages” to come anytime soon.
That’s because CSP and ISP systems require full automation of at least one network domain, “plus a level of comfort with the AI and maturity of automation” to ensure there won’t be any issues without a human in the loop.
“However, AI-assisted operations, where recommendations are made and humans get to evaluate the actions suggested before they are activated are certainly possible and we'll see a lot more of that in the next 12-18 months,” he said.
ISPs and AI woes
We may still be in the early innings of AI. But we can expect to see “a lot more applications” come into play starting in 2025, said Andy Jolls, founder and AI consultant at Roxology.
“To me, last year was the year of content in AI, like let’s go solve content problems,” Jolls said on the FBA webinar. “This year is more, let’s go and solve the workflow and data analytics problem.”
Although not all ISPs may have the expertise to tackle AI, “the majority have somebody in there who’s sort of got a keen interest in this stuff,” Presement said.
Most of the work ISPs are now doing with AI is centered around customer service applications. GenAI co-pilots can help sales reps craft offers and emails to prospects, said de Grimaldo, and, in general, AI can "provide insights on next-best-action to take to support a specific customer."
Implementing AI for network operations (NetOps) is a bit more complicated.
“There’s been a lot of criticism around, is the juice worth the squeeze on the NetOps side?” said Presement. “Are they going to spend the money to make the changes to get some sort of return?”
The “cost of failure” of AI in the NetOps domain is “more wide-ranging,” noted Chua, as one mistake could impact a large number of subscribers.
Many CSPs are still working through what their AI strategy needs to be and adopting policies for how to use it. “Consuming GenAI as part of a productivity suite from Google or Microsoft is quite straightforward, as is enabling GenAI/AI features as part of SaaS packages like SalesForce and ServiceNow or data platforms like Snowflake,” Chua explained.
But providers still need to modernize their data platforms to unlock AI’s full potential, “which usually means some level of cloud architecture transformation.”
“And not all of them are quite there yet,” Chua said. “Plus, the higher cost of capital and capex squeeze makes it harder for them to invest freely in new initiatives today.”
Want to learn more about network automation? Register for our free Cloud Automation Summit virtual event today.