The number out there is $2.5 billion by 2022. That’s the market value forecast MarketsandMarkets provides for artificial intelligence in the telecommunications industry alone.
While the implications of AI and machine learning are vast for telecoms, there are several examples of how telecoms are using their data and machine learning models across their businesses. Here are a few of them, which can be applied for various use cases , whether mobile broadband or fixed telecommunications.
1. Improving customer network experience
Even with a well-developed infrastructure, one if five mobile customers claim to have had a recent bad experience. Intelligent network analysis brings together multiple points of data and provides actionable insights into the factors behind the mobile experience. In fact, the analysis helps telecoms understand which customers actually had a bad experience caused by the network.
By applying data science and building an AI platform, telecoms can evaluate the customer experience, measured by a performance index, and predicts the potential for bad experiences.
However, the model analyzes the experience from the customer’s perspective, providing insights to how the customer is feeling about the experience. Once the data is processed by the model, it learns and provides fewer yet more meaningful alerts on user experience.
Telecoms use the insights from AI-powered analysis to understand which customer experience issues are in their control and what can be done to alleviate the issue. Other issues can be ignored, as they are out of the telecom’s control.
2. Automating customer service
Customers don’t like waiting. So when a mobile user reaches out to a service provider, they want answers… now. A telecom could double or triple the size of their customer service departments to achieve faster response & answer rates, but it’s not a feasible solution, as the telecom would have to hire more people as response times crept up. And a bot can communicate with thousands of people at once — something a human cannot do.
AI models learn why customers reach out to their service providers and can predict when a customer will make a contact, allowing the telecom to take proactive action, or provide an answer now if the customer makes contact first.
In some cases, AI-powered chatbots can automate the replying to simple questions and will remove the need for human intervention for similar or related questions.
Through the use of an AI chatbot, Elisa Estonia has already noted that it’s possible to automate 70% of contacts where the customer needs a fast response, but is a very routine request.
How the AI chatbot model works
AI chatbots can automate customer inquiries, while intelligently forwarding up-sell and cross-sell opportunities, and the need for human intervention through chat can get down to 20% of all user queries.
3. Increasing sales
Speaking of sales, good salespeople strive for a better closing ratio. To achieve a higher closing rate, the appropriate product or solution must be offered to the right customer at the right time.
AI-powered sales tools help telecom sales departments do just that. An AI model built around a telecom’s products, their customers and customer behavior gives salespeople recommendations for next best offers.
A recommendation engine uses a lot more information than a human salesperson can access and process in the midst of a sales conversation.
Add an AI model to the telecom’s CRM and the salesperson can see potential offer recommendations as the customer is pulled up in the CRM software.
4. Network infrastructure optimization
Now onto network infrastructure — software-defined networks have led to an abundance of data on the network itself, allowing AI models to learn, optimize, and self-heal. Data about the network is processed in real-time, allow telecoms to detect potential hardware issues before they get serious, avoiding network outages or congestions that customers despise.
The AI model learns when additional load-balancing may be needed, or when a system needs to be restarted. Fault prediction alerts engineers when the hardware is about to malfunction, allowing technicians in the field to alter configurations before it actually fails.
In addition, quality control through image analysis can detect if a field technician has missed something. AT&T takes image analysis further by planning to use drones to capture video of cell towers, which the algorithms process to understand if there are any issues with the infrastructure.
5. Revenue assurance
Saying that billing in telecommunications is complex is an understatement. Revenue data comes from dozens of databases, about millions of customers, and through potentially hundreds of partners. While logic is applied to the databases, if something breaks, or there’s an incident of fraud, the company could lose thousands to millions of euros (or dollars.)
Anomaly detection algorithms find deviances from usual in the trends. The model then alerts the experts, who evaluate whether the deviation is expected, for reasons like changes in the pricing policy or pricing specials, or there’s an issue with the system, or there’s potential for fraud.
As the algorithm runs and the experts provide feedback, it learns and gets better at understanding the anomalies and detecting fraud.
Kristjan has been studying and working on machine learning projects for more than 5 years.
After acquiring a Master’s Degree in Computer Science and Machine Learning, he started working at Milrem Robotics as the Team Lead for Autonomous Vehicles, helping to build self-driving vehicles. Kristjan also has experience in building intelligent systems for data centers, robots and electric formulas; and with computer vision and image recognition. He is especially fascinated by how people from different industries combining their knowledge with data science arrive at new insights and help to accelerate innovation.