While the algorithms utilized in artificial intelligence are widely available to telcos, the critical challenge, according to Niilo Fredrikson, the executive vice president of Comptel's Intelligent Data Business Unit, is to apply these algorithms shrewdly.
Fredrikson told Telco Transformation that the core of the task is to develop platforms that use these powerful AI and machine learning capabilities to more intelligently address prospects at the right time and with a unique and personalized message based on that person's known tendencies and prior behavior. Late last year, Comptel introduced Fastermind, an AI-based platform aimed at telcos that can be utilized to improve customer engagement.
Telco Transformation: What is Fastermind and where does it fit into Comptel's overall organization?
Niilo Fredrikson: Comptel Corp. (Nasdaq, Helsinki: CTL1V)'s business is structured into two business units. The first is called service orchestration. The second, which I lead, is called intelligent data. Our focus is the digital customer journey.
This means, essentially, enabling telecom operators to create smooth contextual and intelligent experiences for their end customers across all the touch points, including the cross-sell and upsell offers and customer care inbound and outbound touchpoints. Through these, telcos will be able to increase revenues, increase customers' satisfaction, and also decrease time to market.
We make it happen is with three product portfolios. The first one is called Data Refinery, which is a data integration and complex event processing platform. The second product is called Monetizer, which is an agile offer creation tool that sits on top of policy and charts and functionality enabling the alteration and deployment of new data plans in minutes.
The third product is Fastermind, which includes two core functionalities. It's a real-time decisioning logic builder and a rules engine. The second part is the AI part, which is essentially a set of deep analytics and machine learning capabilities.
TT: What is the first element?
NF: There's something we call Real-Time Decisioning or RTD. That essentially is a business logic builder. For example, a product manager or a marketing director at a mobile operator can use this user interface to in a graphical way to build different customer engagement logics, which are then automated and which leverage the second piece, which are the analytics and machine learning capabilities.
TT: Can you provide an example of how Fastermind works?
NF: Let's look at the subset of those customers who either through their phone or through a computer added an iPhone into their shopping basket but have not completed the purchase. Then let's add a location trigger, which says that, okay, a customer, who fulfills these two criteria comes close to our retail location. Let's send him a notification in real time, prompting the customer to visit the retail location and say, "Hey, we have the phone that you were looking at in stock. Why don't you come by and we can show it to you?"
Then there is a fourth piece, which is now related to a deep analytics part. Let's also create a special discounted offer for this particular situation or customer based on an analytic score, which is an estimate of that customer's likelihood to react to different size discounts and make a purchase decision based on that.
TT: This sounds like it can be done with a sophisticated rules-based engine. What is the AI element?
NF: In this specific example, the true AI part of it is the machine learning aspect of determining the customer's likelihood to buy based on a discount offer. It is about the analytics score, which tells what the customer's propensity to buy is based on a discount.
Then, there is a feedback loop based on machine learning where we look at all the purchases and new data coming in and constantly change the algorithm so that it learns by itself and becomes better and better. The score is individual for each user. The algorithm used is a self-learning neural network, which is by nature a machine learning algorithm. It adjusts the results and outcomes based on data coming in over time.
Next page: Two approaches to AI: number crunching and 'thinking'
TT: There seem to be two approaches to AI: Number crunching and "thinking."
NF: That is accurate. In its simplest form, you can essentially divide technologies into these two types by looking at the kind of algorithms they use. When talking about neural networks, which are typically used in machine learning, people sometimes use the term "deep learning." That is the human mind-like approach. The other approach, big data crunching, is based on more traditional computation, the horse power-type application of analytics algorithms.
TT: Which does Fastermind use?
NF: Both. We've combined them in Fastermind, very consciously. There are lots of engines and products and platforms out there that do exactly those kinds of things. There is nothing magic about it. The base technology and algorithms are out there. While it's obviously a technical challenge, there is no super-secret hidden sauce.
The secret sauce is how you apply the technology. That's why we've included in Fastermind the business users' tool, which we call the Logic Builder, which you can use in a very specific mobile operator context for leveraging the AI capabilities and connecting them to customer engagement work flows.
TT: So there's an interface to the sales channels, CRM platform and big data repository with all the data from customers, so on and so forth. This creates a bridge to tap into that and ride on top of it.
NF: Exactly. It's the ability to [create] extra data for the user of the analytics and AI capabilities. But is also works the other way around. Based on the results, actions are pushed into various channels. The first action in the example was that the customer received a notification on his or her phone saying, "Hey, why don't you stop by our retail location?"
But it doesn't stop there. At the retail location the agent on his iPad gets a notification saying, "Hey, this customer is coming in. And by the way, we know that if we offer a 15% discount to this particular customer on the iPhone, there's a very high likelihood of closing the sale."
TT: When was Fastermind introduced?
NF: We launched the Fastermind suite in its current form at the very end of last year. Salesforce.com Inc. , who we consider a partner, has included us in their telco blueprint as the recommendations engine. In other cases, we offer it directly to mobile operators and then integrate to the systems that they have in place.
TT: What is the value of contextual and intelligent customer engagement?
NF: Look at how operators engage with their customers. Their average net promoter scores, their net satisfaction scores, are as an industry horrible compared with pretty much anybody else, and especially if you compare them to newcomers such as WhatsApp and Amazon.
What we've noticed is that [there are big improvements] when customer engagement is personalized and contextual and intelligent, meaning that I as an operator offer you something which makes sense. That's an intelligence piece. If I offer you an iPhone but you want an Android, it doesn't make sense. That's the first check point: Does it make sense?
The second is that it needs to be personalized. It needs to be part of what you've been doing. It makes sense personally for you based on other actions you've taken previously.
Then the last is contextual. If you happened to be having dinner with your family, that is not a great time for me to offer you anything, even if it is free, because you don't want to be interrupted. However, if you happen to be commuting to work on a train, that could be a great time to communicate with you.
— Carl Weinschenk, Contributing Writer, Telco Transformation