Why is streaming data and real-time AI critical in telecom?
In an era where connectivity is the lifeblood of our digital world, the telecom industry stands at the forefront of technological evolution. As the demand for seamless, fast communication continues to skyrocket, the integration of real-time AI and streaming data has become more than a competitive advantage — it’s a necessity.
The Future of AI in Telco
Recently, I listened to a very insightful podcast interview about “The Future of AI in Telco” with Mazin Gilbert from Google. Before joining Google, Mazin had been working 25 years at AT&T, and 6 years at AT&T Bell Labs, so he knows quite a lot about telecoms. You can listen the whole podcast episode here
Among the others, he said that nowadays customers use high-bandwidth applications e.g. YouTube, TikTok, and Facebook that transfer a lot of data in the form of video, music, images, and files. What’s important is that customers also expect that their mobile network provider will give them a reliable network to consume such content — smoothly, at any time, and for a good price.
From the telecom perspective it means that they need to have a very energy efficient, agile, low-cost network that can scale. This is a must-have requirement. To build and operate such a network, of course, real-time analytics & AI is needed, for example, to monitor its health, ensure its quality, do proactive maintenance, and detect anomalies.
This is actually the project that Networks has implemented and recently presented at an international data conference “Flink Forward”. Their real-time platform processes 2.2 billion messages per day coming from the network management devices of two main mobile operators in Poland (i.e. Orange and T-Mobile) and provides flexible analytical capabilities in real-time. The system processes various metrics coming from 750 data sources, including the calculation of more than 5000 KPIs and 1500 aggregations defined in SQL in real-time. You can watch this presentation about it here:
Streaming Data in Telecoms
Analysis performed on real-time events not only allows you to understand the network health and predict failures but also discover patterns in users’ behavior that can help to better locate 5G antennas. Telecoms run advanced simulations to understand where to put what antenna (e.g. what directions, which building, what traffic it might support) — millions of such simulations can be computed before actually engaging constructors and engineers to install new antennas and equipment. This saves the costs and improves the revenue.
Also Swisscom has presented their network monitoring and incident management with real-time data at scale to inform customers about outages, root cause analysis, and much more. You can find one of their presentations there Swisscom Network Analytics | SwiNOG
On top of that, telecoms can use (real-time) data & AI to offer personalized product recommendations, understand marketing ROI, and provide great customer support with Conversational AI.
A part of this was included in the scope of the project that Kcell (a telecom with 10+ million customers) has implemented and presented at “Big Data Tech Warsaw”. Their system runs marketing campaigns, offers new products, and detects fraud by following the behavior of millions of users in real-time and reacting to it instantly. Their system is processing hundreds of thousands of events per second. You can watch this presentation here:
Telkomsel, the world’s 7th-largest mobile operator, runs dozens of machine learning models in their production environment e.g. to better tailor their offers for customers, create and launch new businesses to reach consumers underserved by traditional home broadband infrastructure, simplify the way customers engage with Telkomsel’s services.
There are also use-cases that can be implemented in a batch manner, and they are still good enough e.g. analyzing the cost of network incidents, and churn detection. However, having them analyzed in real-time brings, of course, more value. You can also watch how Kcell implemented them from the video recording from one of the conferences:
From 20% to 80% on Real-Time AI?
An interesting insight shared by Mazin Gilbert is his estimation that telecoms currently spend 80% on data management and 20% on added AI capabilities. He thinks that in 5 years, it will be the other way around, so 20% of time will be spent on data management and 80% on building value-added AI capabilities. Of course, you will never spend 0% time on data management, because AI will not do everything for you, but AI will be increasingly adopted for more and more use-cases.
In my opinion, in the telecom industry, most likely the data that you will spend these ~20% managing will be mainly streaming data, and 80% of AI that you will use will be mainly real-time AI/ML.
If this topic is interesting to you, you can also read more in “The State of Data Streaming for Telco in 2023”.
Would you like to know more about streaming data analytics or real-time AI? If you want to discuss this or any other topics, feel free to sign up for a 30 minutes free consultation with our experts.
Blog Author: Adam Kawa — CEO and Founder
Originally published at https://getindata.com.
