The five things telcos need to know about implementing AI
Artificial intelligence has the potential to dramatically transform the telecoms industry. Over time, AI will provide telecom operators with the insights and capabilities they need to automate and proactively address issues in customer care, service operations, security operations, and network operations.
But when it comes to AI implementation, we’re still very much in the early days. While there’s a lot of hype around AI, there’s also a lot of confusion, with many operators unclear about what AI can do for them — or how to get their AI initiatives off the ground.
To cut through the hype, Nokia commissioned a research study on the current state of AI in the telecoms industry. Based on a global online survey of AI leads at over 50 telcos as well as nine in-depth interviews with people in roles including head of predictive analytics, here are five things every telco needs to know about implementing AI.
1. AI isn’t a standalone technology, it’s a journey.
Broadly speaking, AI refers to the computing capabilities that allow machines to perform human-like functions, such as inferring and planning, in human-like ways. But there’s no single device or application that “does AI” — it encompasses a broad spectrum of concepts and technologies. And to be effective, it needs to be embedded in all aspects of a CSP’s business.
Today’s telcos are starting to take steps along the path. Many are using targeted data analysis to inform human-led decision-making or even fixed-policy automation where machines follow scripts written by humans. As a next step, telcos should aspire to have automation supported by machine learning, with systems and processes that run, learn and self-improve with limited human input. That will move them further along the way to one of the major opportunities with AI: fully autonomous systems.
2. Telcos are looking to AI to improve current processes.
The telcos we surveyed said AI will have the biggest impact in three areas: optimizing networks and operations, augmenting sales and marketing, and enhancing the customer experience. The common thread? They intend to use AI to improve existing business processes.
That’s a good starting point. But telcos need to play the long game, too. That means thinking about how AI can unlock new revenue streams by monetizing customer data or launching and supporting new services.
3. AI implementations come in three main flavors.
Nearly one-quarter of the people we surveyed said their company is taking a centralized approach to AI implementation, where a senior executive leads the initiative to ensure it stays at the top of the agenda. Others have set up cross-functional R&D units to serve as “internal vendors” to business units, using AI to solve specific business problems. And some are leaving it to the business units themselves to run their own isolated initiatives.
Our research shows that telcos who adopt some level of centralized coordination are much more likely to convert pilot AI projects into live or scaled projects. Of those taking a centralized approach, 82% have at least one live project compared to just 38% of those running siloed initiatives.
4. Telcos are in a good position — if they can handle their data.
AI requires vast quantities of data to function correctly. The good news is that telcos arguably have more subscriber information than companies in most other industries. But that data must also be of sufficient quality — and that’s where many telcos struggle.
The data they have isn’t “clean” enough to feed into machine learning algorithms, with 56% of telcos facing problems with inconsistent or fragmented data. And more than three-quarters said their data storage systems need to be improved so that the information collected can be labelled and organized in a way AI systems can use. A more disciplined approach to data is needed — one that also accounts for data privacy and security.
5. New partnership approaches are needed.
Very few of the AI projects in our study made it to live deployment. The industry is at risk of getting stuck in “proof-of-concept purgatory” — a limbo state where AI stalls out and never reaches its full potential.
Because most telcos don’t have the resources to fully develop or implement AI on their own, some level of partnering is necessary. But the traditional way of doing that — rigid procurement processes, buying ready-made solutions — no longer works. Only 29% of telcos who stuck to that traditional approach managed to go live with AI versus 61% of those who explored ways of working more collaboratively with vendors and branched out beyond their usual group of vendors.
The pressure is already mounting for telcos to go “all in” with AI. As customers continue to expect more dynamic, responsive and personalized services, AI will be essential to anticipating and accommodating their needs faster and in a more agile way. Telcos who start their AI journey now will put themselves ahead of the pack.
Want to learn more about AI in the telecoms industry? Download our executive briefing to get the full details.
Share your thoughts on this topic by joining the Twitter discussion with @nokianetworks or @nokia using #CSPs and #AI.