A Conversation with Lester Thomas from Vodafone Group
Part 1 of 2
Introduction
We sat down with Lester Thomas, Chief IT Systems Architect at Vodafone Group, and asked him a series of questions about his experiences with, and opinion on the use of AI, Machine Learning and analytics in Telecoms.
What impacts have AI and Machine Learning had on modern organisations?
When we’re doing things like network planning, we spend lots of money on rolling out new networks. How do we optimise where we’re investing the money? You want to optimise for coverage where you’ve got customers and where the impact on the customers is going to be the greatest – you can use AI in your operations. We get a huge number of alarms from our network and IT systems, AI has a big role to play in how you automatically manage some of those alarms – we’re proactively removing a lot of the alarms and optimising for that.
When we’re marketing to customers, we want to market as a one-to-one conversation. So, for every customer you have an individual conversation based on the context you’re in and also the context of that customer. We try not to just blast out marketing messages to everyone – the best place to have conversation is when someone’s interacting with Vodafone, can you then put contextual optimised conversation topics in front of them to then engage with them properly?
Often, the thing that the end user wants isn’t a connected device, or even just data from a device: they want analytics. The reason you’re putting something into the world to capture data is to drive the analytics from it – we’re having services where we’re not selling connected devices, or even the data from those connected devices, we’re selling the analytics you can derive over the top of that data. I think a lot of the new revenue streams and new service models for Telcos will actually be based on AI and analytics.
“You want to optimise for coverage where you’ve got customers and where the impact on the customers is going to be the greatest”
How can Telecom companies use AI and Machine Learning to improve customer services?
We’re using AI in all of our digital interactions now, I’ll give you some examples: People are tending to contact Vodafone through text and chat rather than through voice, most of our customers are finding it more convenient to contact through text and chat, with that you actually have more opportunity to use data and analytics. We have an automated chat bot we call Toby: If you contact Vodafone through our web chat, across all our markets, your first interaction will be with the robot. The idea is that most common interactions we’ll do completely automatically, and much quicker than waiting for a human agent at the other end. In the conversation, if it’s something that the chat bot doesn’t understand, you get passed forward – but you’re getting passed forward to the right agent for that sort of query. By then it’s discovered at least the area of the challenge that you’re facing.
The other thing we’re working on is “how do you remove the need for people to contact Vodafone because of a problem?” So, one of the metrics we measure isn’t even first-time resolution, it’s “how many contacts per month, per customer are we actually getting into Vodafone which are caused by unnecessary problems?” You do that by analysing data about the problems people are having, doing a root cause analysis and finding the root cause for those problems, or if something goes wrong in our network – can you use this automation to fix the problem proactively before it causes a customer to contact us?
How can Telecom companies use AI and Machine Learning for intelligent network provisioning?
You can have automated algorithms which can optimise a solution: Lots of network planning, things like that, you’d call them an optimisation challenge – you can use AI for what we call complex decision-making. There might be stages in your operational processes where historically you might have gone to a human expert to make a decision, but it’s much better to encode all that expertise into a repeatable pattern. In your complex decision-making, you come to a point and say “I’m going to go to this expert system where I can use my big data” and make some complex decision. You have the example of AI agent-triggering actions: you can have an agent sitting in the background, constantly looking after your customers and spotting when there are patterns, then automatically driving some action off the back of those patterns.
We have a data plane and control plane in the networks, some people talk about this as being like a knowledge plane for a network, the notion that even the network topology can dynamically change off the basis of some intelligence in the network identifying a pattern. We deal with examples where you change the network topology, in the case of a denial of service attack, in normal operations you might have one topology which is like a layer of firewalls, for example: If you spot a denial of service attack, your knowledge plane of your network can then say “I’m going to deploy an alternative topology with an extra layer of high-performance denial of service firewalls to filter out that attack”, so that your normal traffic can get through uninterrupted.
The last one drives the root cause analysis: how do you bring insight and structure to data? You have operations in the manual world where we’re swamped with too much data, but you can use techniques to bring insight and structure to data to identify patterns, so that you can do a root cause analysis and fix problems before they become a problem for the customer.
Check out part 2 of our interview with Lester Thomas, in which he tells us what he thinks we can expect from AI in Telecoms in the near future.
Our partner Bruhati have a successful track record in helping organisations through their Architecture Maturity Journey and enabling them to become fully digital. For more information and a FREE consultation please visit www.bruhati.com or contact sales@bruhati.com
Feel free to comment and subscribe to be notified when a new article is published.
