ML and AI – the future of revenue assurance and fraud management
Machine learning (ML) and artificial intelligence (AI) – algorithms that learn from data and make predictions – are unlocking opportunities for businesses in almost every field. If you’ve shopped online and suddenly seen advertisements for items suspiciously tailored to your taste, then you’ve seen the wonders of ML and AI at work.
In the telecoms industry, ML and AI show particular promise for revenue assurance and fraud management (RA/FM). Each year, telecom companies lose billions to fraud and revenue leakage – money that falls by the wayside because of human or computer error. RA/FM teams are tasked with detecting and preventing such unnecessary losses.
Dr. Gadi Solotorevsky, a revenue protection expert and an AI veteran, discusses ML and AI and explains why algorithms will assist revenue assurance (RA) and fraud management (FM) teams, but never entirely replace the human element.
VanillaPlus: How have AI and ML changed in the past five or ten years?
Dr. Gadi Solotorevsky: Siri, Alexa and Cortana once sounded like exotic names. Today these intelligent personal assistants are part of everyday life. They’re living proof that AI and ML are useful. Things considered science fiction when I was a computer science student in the 80s – like the idea of computers playing chess – are now mundane. We don’t even think of a smartphone playing chess as AI or ML, but it is. As the technology matures and becomes more accessible, we’ll be able to benefit from them more.
VP: What triggered usage of ML and AI in the RA/FM space?
GS: RA/FM involves detecting patterns, outliers and learning behaviours – all while dealing with massive amounts of data. This is a field in which AI and ML can really excel. But it is not enough to have good technology. Change also requires effort and adaptation. Users need compelling reasons to adopt new technology. I believe these currently exist in the field of RA/FM.
First, telecommunications are changing drastically. In the past, the industry included a small number of relatively static products – such as voice and cellular. It changed drastically. Today there’s a proliferation of products – smart home, telematics, eHealth – and an emerging application programme interface (API) based economy which encompasses billing as a service, for example. All of it is empowered by a digital ecosystem that allows the creation of new services or products on the fly. RA/FM professionals have to find a way to keep up with this pace.
Second, after two decades of modern RA/FM there are still significant leakages. The last RA report published by TMForum estimates communications service providers (CSPs) lost about 1.5% of their total revenue due to revenue leakages. Something needs to be done to reduce this. AI and ML can come to the rescue by detecting and preventing fraud and revenue loss, helping keep pace with the rhythm of innovation in products, services, and business models of the service providers.
Using AI and ML in RA/FM isn’t a thing of the future. It’s happening here and now, and we’ll see more embedding and integrating in RA/MF products and methodologies.
VP: Will ML and AI eventually replace traditional RA/FM systems and analysts?
GS: Machine learning is not meant to replace human RA/FM specialists. Rather, to assist them. In the near future, I predict data scientists will become key components in RA/FM teams.
I’ll explain further. Some people think you can throw a problem at an AI and ML system and it will magically resolve it. That’s naïve. Deep Blue defeated Kasparov in chess in the 90s, but it took two decades for AlphaGo to beat the world’s best GO player. The two problems are relatively similar, two board games for two players, yet there was a 20 year gap between solving them.
Similarly, solving RA/FM problems requires using and adapting the right AI and ML techniques for the right time. Service providers have similarities and fraud techniques often resemble one another – yet neither are identical. Developing a completely autonomous AI and ML system for RA/FM is impractical using today’s technology. Instead, we should focus on using AI and ML techniques to assist human experts and relieve them of some of their workload – and we’re already doing it.
The role of data scientists in the RA/FM team will be to create and adapt the ML toolset to the realities they face. AI and ML will be used both independently to detect new issues, and in combination with the traditional rule based techniques. For example, it can be used to tune parameters or reduce the number of false positives. RA/FM analysts will continue to investigate and root out possible problems. AI and ML will help them better investigate new directions, increase the RA/FM protection coverage, and improve accuracy.
VP: We hear identity theft and impersonation are key enablers in fraud – is it possible to use AI and ML to detect and prevent them?
GS: If you prevent impersonation, you can prevent lots of fraud. Impersonation not only opens the door to traditional fraud such as equipment theft, but far more grave threats once you consider new services like smart homes and telematics. Verifying customer identity is crucial.
One approach is to apply endless layers of active protection. Good examples include entering a code, using fingerprint or retinal scans. To access my email on my phone I unlock the phone using my fingerprint, then I have to enter a code to enter the email app. regretfully, these methods are tedious. Worse, they can still sometimes be hacked.
The old saying: “If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck,” is relevant. You can learn behaviour patterns of users and use that data to detect would-be hackers. It’s easier said than done, and of course there are some privacy issues, but AI and ML can do this magic. We can use AI and ML to learn behaviour patterns and then use these patterns to detect imposters.
These are exactly the kind of applications that we’re working on today, and they will be part of future of RA/FM solutions.
IN ASSOCIATION WITH AMDOCS
Amdocs Revenue Guard –
an industry-leading risk assurance management – recently won an award at the TMForum in Nice for participating in the catalyst Connected Citizen: Life in a Green, Clean, Smart City. Amdocs contributed advanced analytics, cyber-fraud and behavior analysis to the project championed by Orange Group and NTT, in collaboration with BearingPoint/Infonova, ESRI and Symantec. Revenue Guard’s role addressed key challenges of the smart city/IoT ecosystem by verifying identities, preventing equipment theft, unauthorized usage, e-health monitoring and cyber-protection against unauthorized access. It currently has a patent pending for the technology it used the catalyst.