Broadband World Forum (BBWF): At Broadband World Forum 2019 we are cutting through the hype surrounding technologies like AI in the industry. There is much debate today about true AI vs. enhance machine learning. What are you seeing in the industry today?

Jay Shah (JS): A lot of people keep using this term ‘AI’ and it’s one of the most annoying things going on now. They’re not really using AI, they’re using Big Data modelling but they’re just marketing it in a fancy way. Real AI has the ability to analyse behaviour, learn trends, and informs your decision making. The broadband industry typically uses what they call AI for network optimisation and optimising your marketing strategies. However, there is so much more it can do for businesses.

BBWF: Can you expand on these two uses?

JS: When it comes to network optimisation, organisations can look at all the data that is being exchanged on a network – for instance when an individual is gaming and what game they are playing - but they can’t see what the user is doing in the game. What organizations can do, however, is optimise the interconnects between their own servers and the gaming servers. From the gamers perspective they experience the lowest latency when using that internet access; and when a millisecond makes the difference between life and death in a game this is important!

From the marketing perspective, churn management is a huge problem in the telco industry. Identifying churn is a long-winded and complex process especially since it relies on analysing mass amount of data about your subscribers. At MyRepublic one of the approaches we are using us plotting all subscriber information on a map. For example, in Singapore the map is broken down into grids, each grid has many zip codes and different target audiences as well as different associated lifestyle and activities. This information together with the existing subscriber data is used to profile the customer as well as predict the likelihood of churn. These predictions are then sent to the relevant departments to manage the process.

We also use this map information and data flows to optimize user acquisition. In the last three months alone we have dropped user acquisition costs by as much as 84%.

BBWF: Is this what you are seeing telcos do today?

JS: The traditional ways they use it is by saying they harness artificial intelligence for customer services or network optimisation, and they use buzzwords like self-healing networks. In reality, they are not AI models but instead data analytics.

True artificial intelligence will analyse patterns, use them to make recommendations to consumers and, depending on how they interact, develop a feedback loop in the system and learn from its findings. In MyRepublic, for example, this helps with predictive decision making, informs our decisions when developing new products and services and identifies new revenue lines for the business.

BBWF: Looking forward, how do you think artificial intelligence will continue to evolve the industry?

JS: We’ll use something we’re building in MyRepublic now as a demonstration: a true profiling and recommendation system. We use internal data to map and analyse a variety of different factors that contribute to decision makers, such as location, time of day and lifestyle choices, and using these to build extremely detailed profiles of consumers. These profiles then inform our marketing campaigns, retention strategies, style of out-of-home advertising and so on, making them far more personalized and effective. We’re also building an AI-powered chatbot that looks at how people communicate with the bot, analysing information such as tone and speed of writing. The bot then continuously learns from this information to produce better and more appropriate responses, a stark contrast to the majority of chatbots out there that are glorified knowledge bases. Telco services will continuously become smarter, more personalised and faster as true AI is implemented – where applicable.

BBWF: What are the quick wins for operators deploying network AI, in your opinion?

JS: There are no quick wins – you cannot build an AI model in a few days and call that a win. My advice would be for businesses to firstly identify what their core objectives are and how they need to get there and then determine whether they truly need AI, or do they simply need data modelling.

BBWF: Finally, how can organisations go about finding the right solution for their business?

JS: In the telecom industry it’s one of the toughest things to do because margins are so thin; costs are going up and your wholesale costs aren’t going down, and yet your retail price is plummeting. You have to optimise everything and every layer in the company but don’t have the capacity or budgets to do so.

It’s an interesting landscape because you have tech companies like Rakuten moving into the mobility space, and MyRepublic who are a telecoms company who are moving into a tech space, so organisations really need to shift their mindset towards functioning more like a tech company and utilising all their assets. Ultimately, it is about survival and then adapting to changes in the marketplace to identify new revenue lines for the business.