Accelerating AI in Telecommunications

We recently attended the Telco AI summit in London. Federico Castanedo spoke on a panel about redefining network operations using AI, whilst I delivered the DataRobot keynote on accelerating AI adoption in the industry more generally. We also spoke one-on-one with over fifty telecommunication (telco) decision-makers and had the opportunity to hear other speakers at the event. The central theme across these interactions was that of acceleration. Telcos are accelerating their adoption of artificial intelligence, but need help to further accelerate their efforts. Those that win this AI race will reap huge first-mover advantages.

Telcos Take on AI

A survey from McKinsey shows that telecommunications is one of the most AI mature industries. Adoption has historically been focused on service operations, product development, marketing, and sales. Moreover, telcos are amongst the most committed to furthering their investments.

However, like many businesses, telcos have found AI difficult to adopt. The vast majority experienced challenges in building AI applications – it often takes many months or even years to build a model – by which time the data is no longer up to date. And, once they’ve built their AI, most struggle to deploy it into production. Finally, if they can overcome these hurdles, there is still a minuscule chance that their AI is even monitored.

This painful implementation cycle is exacerbated by the scarcity of data science skills and of those who also have a good understanding of the domain-specific challenges of telco businesses. Many of the good data scientists I caught up with have revolved through multiple different telcos over the last five years.

Transforming a Telco with AI

We focused on these challenges, the technological and organizational capabilities needed to overcome them, and the use cases to consider in transforming a telco with AI. Opportunity areas include:

• Vision: Setting a transformation vision for the use of AI.
• Automation: Accelerating as much as possible of the AI lifecycle, from initial use case discovery through to feature engineering, model development, and deployment.
• Education: Pragmatic cross-telco education on AI, not on the latest python libraries but on identifying opportunities and framing projects for success. 
• Productivity: Boosting the productivity of data science teams through automation and prioritizing their efforts in alignment with business objectives over completing “moonshots”.
• Democratization: Empowering business analysts and operations people to become citizen data scientists. 
• Management: Monitoring and managing models effectively in production. 
• Breadth: Using a wide range of algorithms and modeling approaches, not being limited to the easiest or favorites. 
• Deployment: Deploying on the cloud for faster and cheaper implementations, whilst retaining on-premise for more data sensitive use cases. 

As for use cases, we discussed the success of Sky’s DataRobot project, optimizing the delivery and storage of content. Across the conference, there was the greatest interest in applying AI in two broad areas: Customers (a.k.a Customer Value Management) and Infrastructure/Networks.  

In the former, a popular use case was improving customer retention by predicting which customers might churn and then making proactive interventions tailored to their needs. In the latter, many welcomed the opportunity to better manage the maintenance of complex telco infrastructure, reading diagnostic data, and using it to predict faults ahead of time. 

AI use cases in Telco are relevant across lines of business:

• Customer acquisition
• Customer care
• Customer retention (e.g. customer churn predictions)
• Forecasting / Planning
• Human resources
• Marketing
• Network
• Infrastructure
• Operations
• Personalization
• Store operations
• Store planning

If you too are interested in accelerating your adoption of AI, DataRobot can help. Schedule a demo.

This is a repost of an article I wrote for DataRobot here