Data Analytics & AI

Making the business case for investing in AI

At TM Forum Live! Asia today, I spoke to Dr. Rujikorn Pavasuthipaisit, Director of Data Analytics and Research, True Corporation, in a fascinating ‘fireside chat’ focusing on how telcos can make the business case for investing in artificial intelligence (AI).

Of course we are all aware of the hype around AI. And, while many excitable companies go in all guns blazing with little thought of what they want to achieve, others are so wary about what they see as the unknown that they want nothing to do with it.

In this discussion, Dr. Pavasuthipaisit draws on his AI experience, particularly looking at big data, predictive models and customer insights. He discusses how telcos can evolve with the help of AI, how they can go about getting AI smart and some of the things that are stopping them getting there.

ABP: Tell us a little bit about True Corporation, and what it’s doing in AI.

RP: True Corporation is an integrated telecoms provider in the areas of mobile business, broadband and pay-TV, and we use AI in all three of these areas. We use it to predict churn rate, advise customers what services they might need in the future, for upselling and cross-selling, and for figuring out which customers would buy more products from us.

The other position we take with AI is on the customer care side of things. On a daily basis, we bring up alerts on the levels of customer dissatisfaction and the reasons behind it and then take the relevant action. We’ve also been taking it one step further, using deep learning to analyze not only our customers’ needs, but also using AI to analyze their voice and their tone to understand everything they’re not telling us outright.

ABP: We’ve heard a lot about data’s crucial involvement in AI; how much of a challenge is there in utilizing data to enable AI?

RP: It really depends on your business objective. I have two main examples of how data can be used. The first is expending low-level analytics by using business intelligence to create a reporting dashboard. This can help the company establish the necessary data structure for AI.

The second is building data models to develop AI; they develop the model on their data site. So I’ll go back to my example use case of detecting customer dissatisfaction through the customer’s own voice. Examples of satisfaction and dissatisfaction are input into the AI system, which is ultimately what the AI learns from, and what we call supervised learning.

The thing is, this kind of learning doesn’t have predictability, which is where AI eventually needs to be.

ABP: Is this still a challenge? How much data do you need to enable these machines to improve?

The challenge is to find the right kind of data – the right kind of unstructured data to be exact. For this example, it would be a voice or video recording. You’d also need to try and change the framework of your analysis from supervised learning to unsupervised. That can help, and that is the future of AI, a future where AI learns from its own knowledge and the knowledge of other AI. Also, the more relevant data you have, the better as it allows more scope for learning and improvement.

ABP: So you said that you use AI to help with things like predicting churn and helping with customer care. People want to understand what kind of return on investment you can see, as it is expensive to start investing in AI. How has your use of AI helped your company and helped reduce its churn rate?

RP: When we’ve used AI proactively for retention, the churn rate dropped by 10 percent, and that’s only the tip of the iceberg. But, like with customer care, there are a lot of things you might not be able to evaluate in terms of monetary value.

When it comes to developing the business, we used the telecoms data framework and what we learned from it, and applied it to other business ventures. For example, in a medical school in Thailand, we used AI to predict the heart attack rate.

ABP: That’s an interesting point: find new ways of making money. I’ve been in telecoms for about 11 years, and we hear a lot about telco being a “dumb pipe”. Do you think AI is a way to digitally transform and get out of this scenario of being just a dumb pipe? Like you said, it’s opening new revenue streams, like being in medicine, telematics, etc.

RP: Most certainly, yes. But, the disadvantage for telcos compared to new digital companies is the way their operations are set up. It makes it difficult for them to digitalize or utilize existing AI. These kinds of legacy operational models and business models might have been healthy in the past, in the world of first age mobile, but right now, if you want to compete with the giants, it’s difficult based on their new business models and telcos’ old ones.

What most telcos need to do/have done is a spin off [create a separate business entity, away from existing operations, to nurture AI growth]. That helps.

It remains to be seen how telcos would compete in this new digital age. Yes, of course, AI is coming, and it will be the decisive battlefield for global corporations.

ABP: I heard a presentation recently that said for telcos to set themselves up in AI, they shouldn’t think like a telco, but like a new startup.  Do you think telcos can do this, or are cultural problems an issue in terms of developing new areas in AI?

RP: Telcos have a cultural structure which is suitable to help them to be a successful telco, but if they want to develop successful AI, that’s a whole new thing that they definitely need to address.

ABP: You’ve talked about developing new areas of business to help do that, like medicine. How do you get started with this? Which areas do you decide to invest in? We’ve seen telcos try and fail in a number of areas. How are you making a success of AI?

RP: The answer to this question has two levels to it. The first is the development of the model – you have a lot of open source tools, and some companies might even want to resort to cloud services to develop.

The second is model implementation. If you want to implement your model commercially, you need to invest, particularly in big data. You need readily available data, to be able to act fast. A data lake, alongside cloud services will allow you implement business solutions fast.

I’d say another one of the big areas for investment is the people side, which is what the tech giants do. The first step is to acquire lot of highly talented people. In the US right now, it is said that an AI data scientist is more popular than a quarterback. When you have a good team, and the right kind of team (each company has specific needs), it’s a good investment.

ABP: How do you attract those kinds of people? Telcos aren’t seen as sexy as the Googles, Facebooks and startups of the world. That’s the dilemma really. Firms like Google are seen as much better to work for.

RP: If you want those kinds of people, it’s very difficult to compete with tech giants. Telcos don’t necessarily need to attract these people. Instead they can look to the operational frameworks of these tech giants. Instead of creating their own new frameworks, they can use these tech giants’ frameworks and apply them to their own businesses. That’s what a lot of Chinese companies are doing, and they’re doing it successfully.

ABP: Don’t you think that creates a self-fulfilling dilemma? That telcos will always get left behind if they don’t attract the best talent? After all, if they don’t, they won’t be leaders, just reactive.

RP: It depends on your business model and what you want to achieve. If it’s a Google/Facebook kind of business model, these types of companies not actually looking to create revenue straight away. They’re trying to create a platform that dominates. But with telcos, they have a huge CapEx [capital expenditure], and they need to create monthly revenue from investment. If your job target is to curate return on investment, you don’t need to be the leaders of the whole world the way Google and Facebook are.

I’d say that’s the problem for most telcos. Anything not generating revenue is just not a priority for shareholders. A good approach is to instead use the technology of the big tech giants and at least try to create an edge over your competitors. Try to compete as a telco provider in your country instead of trying to be the world’s number one.

ABP: You make some interesting points. Amazon, for example, puts a huge amount of money into AI, and is not bothered about ROI. Maybe that’s what the big difference is with them and telcos. Telcos are looking into the short term, while Amazon is playing the long game.

In that sense, the issue with digital transformation, is we that don’t know where the return on investment is. If anyone tries to drum up interest in terms of investing in new technology, it’s difficult to get exec buy-in without those tangible numbers.

It seems like this continues to be a problem in new areas like AI. Those execs are not convinced, and if this mentality continues, telcos will end up being those dumb pipes they fear becoming, continuing to offer voice data and nothing more. OK they’re expanding into TV, but there are exciting opportunities out there, and unless we change that mindset, we’re not going to be able to change the industry at large for the better.


    About The Author

    Vice President, Content, Research and Media - TM Forum

    Aaron Boasman-Patel is responsible for content across TM Forum’s research portfolio, the digital content platform, Inform, as well as leading content and research across the events and Digital Leadership Summits portfolio, including TM Forum Live! This is fed back to the Collaboration team to help drive the Forum’s strategic direction and collaborative program. He has chaired many international conferences and delivered presentations on smart cities, customer centricity, Internet of Everything, industry developments, market trends and strategy. Aaron has worked in the telecoms and associated industries for over ten years. He graduated with an M.A. (Oxon) from the University of Oxford. Aaron specializes in digital transformation, customer centricity, analytics and AI, and digital platforms and ecosystems.

    Leave A Reply

    Back to top