Ciena pitches generative AI for network planning

Despite all of the buzz around generative AI, it is still unclear what its impact will be on the telecom industry in general and on network planning in particular. While there are still unresolved problems such as hallucinations and copyright issues, AI holds the promise to improve network visibility, efficiency and planning, Ciena's CTO Steve Alexander told Light Reading during an interview held over Zoom.

There has been a lot of talk about using generative AI for customer care and possibly even marketing materials, but Alexander highlights another possibility for early applications of generative AI. He points to database reconciliation as an area that could benefit, explaining that organizations often have a number of databases that contain partial information, such as various contact details for clients, that can be consolidated to create what he calls a "source of truth."

This would also include information about what is going on in the infrastructure, which can then be used for further automation of planning and the network itself. Ciena has been working on improving network intelligence through Blue Planet, the unit responsible for automation. It is also developing a software control offering for network management, control and planning (MCP). By establishing what the normal state of the network is, Ciena can more easily detect potentially abnormal events and diagnose them as, for example, misconfigurations or attacks, says Alexander.

Thanks to the significant trend toward convergence, "one network element can do the work of what ten years ago would have been two or three or four different types of network elements," he said. Combined with better visibility, that could help to improve the customer experience.

Things get more interesting when it comes to network planning, says Alexander. If a service provider can see where there may be a spike in demand in future – due to the construction of a new hospital, for example – it can plan to lay fiber in that area, or develop a clearer idea of where to put routers, switches and data centers.

Another potential benefit is improving efficiency in a large-scale network. "That's another place that AI can help […] you can optimize it for lowest absolute cost, you can optimize it for lowest energy consumption, you can start to do things with the infrastructure," Alexander explained.

Visibility is key

To be able to do that, he adds, the network needs to be capable of adapting. With the right kind of infrastructure, it is possible to optimize, for example, the total number of wavelengths or reduce energy consumption. Alexander cites the example of a network service provider catering to a big sporting event – with sufficient AI capabilities, it might be able to handle the venue's needs without additional manual effort.

There are, however, still many unknowns when it comes to how exactly AI will be applied in the telecom industry, and indeed elsewhere. One big and well-documented issue relates to the data generative AI is trained on and the implication this has for licensing and copyright.

This is also why some of the early applications may be "relatively mundane," as Alexander puts it, focusing on data points already owned by the company. This is the case in database reconciliation. "I already own it, I just want to make it all correct, or I want to make it all consistent," he said.

While there has been a lot of experimenting with applying AI to bigger data sets, information visibility is key, he adds. "You can't automate what you can't see."

One other frequently discussed application of generative AI is in coding, although what the impact will be remains unclear, according to Alexander, due partly to the aforementioned concerns about sourcing third-party data and the lack of clarity over the legal implications.

What of monetization opportunities for operators? While initially Alexander expects there to be more of a focus on savings, he reckons the ability to provide lower latency on demand could help with generating additional revenues. If company boards are willing, operators could also offer generative AI to verticals, such as healthcare or education, either by reselling services or operating private networks for those companies.

Generative AI has clearly piqued the telecom industry's interest. A survey of communications services providers (CSPs) by Light Reading's sister company Omdia has shown that 21% of respondents are already using generative AI, while another 20% are engaging with it in testing labs and 56% are investigating possible use cases.

Related posts:

— Tereza Krásová, Associate Editor, Light Reading