AI in customer services: It’s not all about chatbots

Introduction

Internet companies like Google, Apple and Amazon and second tier digital disruptors like Airbnb, Spotify and Netflix have been refining advanced machine learning-enabled analytics to offer increasingly personalised and predictive services. This means consumers increasingly expect a seamless, personalised experience, where service providers anticipate their needs, rather than react to problems after the fact.

If telcos want to regain credibility with consumers, they must develop more personalised and frictionless customer experiences. Many telcos believe that, as for digital native companies, artificial intelligence (AI) technologies can play a crucial role in achieving their goal to reduce costs while improving services.

In this report, we assess the potential value and feasibility of different types of AI applications in customer experience and customer care and outline how telecoms operators should prioritise their efforts. It is based on primary research at STL events, and conversations with telcos, vendors and other industry players. It also builds on previous reports on big data and AI:

Enter your details below to request an extract of the report

Defining AI: An evolution in analytics

In this and following reports, we are using AI as an all-encompassing term for advanced predictive analytics, based on machine learning technologies. Machine learning should be seen as a stage in the evolution of analytics, from basic data analysis, to big data analytics, to fully automated systems where algorithms continually learn from new data to deliver more efficient and effective responses and automated actions.

In our recent report, Big data analytics: Time to up the ante, we emphasized that as telcos have built up their big data programmes, they have gradually replaced siloed legacy infrastructure, where each department has their own databases and data sources, with a horizontally-integrated infrastructure across all departments. This shift enables telcos to run advanced analytics across much broader data sets and use machine learning algorithms to uncover new patterns and correlations across previously segregated data, often described as discovering ‘unknown unknowns’.

Within the field of machine learning, there are three types of algorithms:

  1. Supervised learning: an algorithm is trained on a large set of labelled data (e.g. photos of cats, or examples of spam mail). The algorithm learns to find patterns in the training data, which it then applies to new data samples to predict the correct answer (e.g. one email is spam, another is not). This is good for situations where historical data can predict future events, such as when a customer is likely to churn, or when a location is likely to see a peak in demand for connectivity.
  2. Unsupervised learning: an algorithm is applied to data without labels and with no specific goal, other than to find patterns or structure in the data. This is good for finding new ways to segment customers or detecting outliers, for example fraudulent activity, or an operator’s most and least profitable customers.
  3. Reinforcement learning: an algorithm is given a predefined goal and a set of allowed actions, and is then left to find the most effective way to achieve its goal through trial and error. This is most famously used in gaming, for example by Google’s AlphaGo Zero. Testing different marketing campaigns to achieve the best result is an example of a business application of reinforcement learning.

Deep learning (DL) is an extension of machine learning where there are many more layers in a neural network, allowing the system to work with much larger and more complex data sets, such as images, video, text and audio, in order to identify more subtle patterns.

In most application fields, ML and DL algorithms are not yet at the stage where they can teach themselves what to do, without any guidance and oversight from humans. Most algorithms are also trained for a specific purpose, so their applications are not easily transferrable. But even with these constraints, the benefits of AI for businesses are clear:

  • Self-improving: ML algorithms continuously learn from experience, either from inbuilt rewards systems or guidance from humans
  • High performing: AI can handle huge amounts of data and never sleeps
  • Human-like: it can interact with types of data previously indecipherable by software, enabling it to automate previously uniquely human tasks, e.g. natural language understanding, facial recognition.
  • Wide reaching: AI is suited to automation of both soft (e.g. chatbots, system control) and manual tasks (i.e. robotics)

Contents:

  • Executive Summary
  • Chatbots aren’t an easy win
  • Predictive care: an easier entry point, but with limits
  • So is AI in customer care worth it?
  • Introduction
  • Defining AI: An evolution in analytics
  • STL Partners’ AI Framework: Sizing the opportunity
  • Chatbots: Is it worth the work?
  • What is a chatbot?
  • Most chatbots are not ready to handle customers yet, and vice versa
  • Telenor: Taking an active role in the AI technology revolution
  • T-Mobile Austria (Deutsche Telekom): A more personalised customer experience
  • Recommendations
  • Predictive care & agent assist
  • AT&T is prioritising predictive care
  • Lots of room for improvement in handling customer calls
  • But predictive care can’t solve everything
  • Conclusions

Figures:

  • Figure 1: STL Partners’ AI Framework
  • Figure 2: Defining customer engagement categories
  • Figure 3: Customer experience is telcos’ main current priority with AI
  • Figure 4: The lifecycle of a chatbot/virtual assistant
  • Figure 5: Chatbots are a long way from meeting satisfaction levels of live agents
  • Figure 6: Telecoms operators say they lack the skills to deploy AI
  • Figure 7: Example of a rule-based versus AI-enabled chatbot in financial services
  • Figure 8: DT’s NLU and conversation flow selection and management process
  • Figure 9: Level of customer satisfaction with UK operators’ complaints resolution
  • Figure 10: Top reasons why customers complain to their service providers
  • Figure 11: A clear preference for predictive care
  • Figure 12: Estimate of telco opex breakdown

Enter your details below to request an extract of the report

Big data analytics – Time to up the ante

Introduction

Recent years have seen an explosion in the amount of data being generated by people and devices, thanks to more advanced network infrastructure, widespread adoption of smartphones and related applications, and digital consumer services. With the expansion of the Internet of Things (IoT), the amount of data being captured, stored, searched and analysed will only continue to increase. Such is the volume and variety of the data that it is beyond traditional processing software and is therefore referred to as ‘big data’.

Big data is of a greater magnitude and variety than traditional data, it comes from multiple sources and can be comprised of various formats, generated, stored and utilised in batches and/or in real-time. There is much talk and discussion around big data and analytics and its potential in many sectors, including telecommunications. As Figure 1 shows, analysis of big data can give an improved basis upon which to base human-led and automated decisions by providing better insight and allowing greater understanding of the situation being addressed.

Figure 1: Using Big Data can result in richer data insights

Source: STL Partners

This report analyses how telcos are pursuing big data analytics, and how to be successful in this regard.  This report seeks to answer the following questions:

  • When does data become ‘big’ and why is it an important issue for telcos?
  • What is the current state of telco big data implementations?
  • Who is doing what in terms of intelligent use of data and analytics?
  • How can big data analytics improve internal operational efficiencies?
  • How can big data be used to improve the relationship between telcos and their customers?
  • Where are the greatest revenue opportunities for telcos to employ big data, e.g. B2B, B2C?
  • Which companies are leading the way in enabling telcos to successfully realise big data strategies?
  • What is required in terms of infrastructure, dedicated teams and partners for successful implementation?

This report discusses implementations of big data and examines how the market will develop as telco awareness, understanding and readiness to make use of big data improves.  It provides an overview of the opportunities and use cases that can be realised and recommends what telcos need to do to achieve these.

Contents:

  • Executive Summary
  • Big data analytics is important
  • …but it’s not a quick win
  • …it’s a strategic play that takes commitment
  • How is ‘big data analytics’ different from ‘analytics’?
  • Opportunities for telcos: typically internal then external
  • Market development and trends
  • Challenges and restrictions in practice
  • What makes a successful big data strategy?
  • Next steps
  • Introduction
  • Methodology
  • An overview of big data analytics
  • Volume, variety and velocity – plus veracity and value
  • The significance of big data for telcos and their future strategies
  • Market development and trends
  • Challenges and restrictions
  • Optimisation and efficiency versus data monetisation
  • Telcos’ big data ecosystem
  • Case studies and results 
  • Early results
  • Big data analytics use cases
  • Examples of internal use-cases
  • Examples of external use cases
  • Findings, conclusions and recommendations

Figures:

  • Figure 1: Using Big Data can result in richer data insights
  • Figure 2: The data-centric telco: infusing data to improve efficiency across functions
  • Figure 3: Options for telcos’ big data implementations
  • Figure 4: Telco’s big data partner ecosystem
  • Figure 5: The components of a telco-oriented big data