In Part 1 of this tutorial series, we gave an overview of cognitive computing and provided examples of how cognitive computing is being used to create industry solutions. We also determined a need for a cognitive platform to implement these solutions. The benefits of a cognitive solution include the reuse of components, faster development of the solution, and reduced costs. The key to creating such a platform includes the ability to run multiple use cases on the platform. Therefore, it is critical to identify these use cases and determine how a platform can provide benefits when implemented. In this series, we are focusing on the Telecommunication and Media & Entertainment industries for our use cases. We chose multiple industries to demonstrate that the platform is not industry-specific, although certain components must be industry-specific depending on the use case.

We begin our journey by focusing on the following use cases:

  • Self-service agent – Device Doctor: A self-service agent lets a customer interact with a communications service provider (CSP) to help resolve specific issues. In this tutorial, we focus on a specific issue that the customer has with their device and explain how cognitive computing can help the customer resolve the issue with no human intervention.
  • Network operations agent: A network operations agent enables a network engineer to more effectively troubleshoot issues within the network. Troubleshooting the network can be a complex task and cognitive computing can support the engineer to reduce escalations and more quickly resolve issues.
  • Personalized TV agent: There is a plethora of content available and finding the right content for an individual can be a daunting task. You can use cognitive computing to provide the right content to a customer based on their interests, no matter the device, resulting in improved customer satisfaction.

Note: We use IBM Watson synonymously with cognitive computing throughout rest of the tutorial.

Self-service agent – Device Doctor

Business problem:

A self-service agent lets a customer pose questions on various topics to a CSP. Watson can analyze the customer’s questions and provide a response. Sample question areas include:

  • Billing and payments
  • Service and device troubleshooting
  • Product eligibility and recommendations
  • Channel listing and eligibility
  • Order changes

Our use case focuses on device troubleshooting and device-related issues such as:

  • Setup and basics: Answers setup questions such as “Can I activate my phone online?”, “How do I reduce the number of dropped calls?”, and “How do I reduce my data usage?”
  • Phone and voice: Answers questions that are related to voice and phone such as “How do I check my voicemail from another phone?” and “How do I pair my Bluetooth headset with my phone?”
  • Connectivity: Addresses areas such as “Why should I turn on wifi?” and “When I leave my wifi hotspot area, what happens?”


The benefits of using Watson in a self-service agent to resolve device-related issues include:

  • Increasing contact deflection because deflecting calls from customer service representatives (CSR) to Watson lets CSRs focus on more important issues.
  • Decreasing the average handling time because Watson can respond as soon as the customer contacts the CSR.
  • Improving first contact resolution, with Watson handling most of the first contact issues because most of these issues tend to be relatively simple. More complicated issues or second contact issues can be transferred to a CSR.

Use case:

Let’s explore how a user can interact with a self-service agent to resolve a device-related issue.

cognitive customer experience managment screen shot

In this image, you can see how the interaction might occur with the Device Doctor:

  1. Josh needs help updating the iOS on a device and logs in to the CSP website. Josh asks his question by using natural language.
  2. Watson is able to understand that Josh would like to update the iOS on his device and checks the CRM system to determine the details of the devices Josh owns. Watson notices that Josh has two devices and asks Josh which device he wants to update.
  3. Josh identifies the device.
  4. Watson recognizes that any iOS update should be preceded by a backup and makes this recommendation.
  5. Josh agrees that a backup is needed, and Watson provides the back-up options.

Watson then continues to work with the customer to update the iOS and confirms that the update was successful.

cognitive customer experience managment screen shot continues with update

  1. Watson asks Josh if he would like to update the device.
  2. Josh confirms that he wants to continue with the update.
  3. Watson realizes that the customer has already updated the iOS on one device. Realizing step-by-step instructions are unnecessary, Watson provides all the key steps in one answer.
  4. When complete, Watson confirms that Josh does not need anything else.

Let’s briefly review what Watson has done.

  • Josh is able to interact with Watson by using natural language.
  • Watson understands the questions and is able provide answers.
  • Watson, in some cases, realizes that it needs to interact with other back-end systems to get additional information before the question is answered. For example, in this use case, Watson determines that Josh has multiple phones by looking at the back-end CRM system, which lists the number of phones.
  • Watson answers in a simple response form or by providing a corpus with more details.

We will provide the details on how we achieved this interaction in subsequent tutorials in the series.

Network operations

Business problem:

Resolving networking issues can be a complex problem. Finding the right resources is difficult, and not resolving the issues quickly can have a huge impact on the service provider, especially if negative buzz is generated. Cognitive computing can assist here by enabling a level 1 network engineer to resolve issues faster. In certain cases, the resolution can be fully automated without any human interaction needed so that the engineer can stay focused on more important issues. Due to the complexity of networking problems, we do not focus on specific networking issues. Instead, we focus on how Watson provides the tools to help resolve the issue.


Some of the key benefits of network operations include:

  • Improved effectiveness of network engineers because they can now interact with Watson and get issues resolved
  • Reduced tier 2-3 escalation

Use case:

Let’s consider a situation where a network engineer is attempting to solve a Border Gateway Protocol (BGP) issue.

  • The network engineer can ask various BGP-related questions, as shown in the following image, and Watson responds with answers.
  • The affected system can pass contextual information about the environment to Watson so that the engineer does not have to keep reminding Watson of the system context. Watson can also examine the question being asked and determine related topics.
  • Watson can indicate the probability that the answer to a question is correct. The engineer can rate this response, helping Watson to improve future answers.

Engineer trying to verify if BGP peers are up

  • During the Q&A session, Watson can ask the engineer to run the debug log analysis tool to determine the error code that is returned. The engineer can locate the code and enter it into Watson, as shown below. In doing so, Watson not only provides information about the error code available in the error manuals, but also related tickets and other relevant data. The following figure shows the trouble tickets that are associated with the identified error code in the context, including environment and previous questions asked. Before this stage, trouble tickets were analyzed, tagged, and learned by Watson so that the appropriate tickets can be provided to the engineer during this Q&A session.

Related tickets screen shot

  • A fairly common problem that network engineers face is prioritizing which issue to work on. The engineer receives alerts from various systems and might have difficulty identifying which one to address first. The engineer might, for example, be viewing a dashboard that lists alerts on certain cellular nodes together with revenue impact if the issue persists at the node.

Screen shot showing several issues being identified

The engineer can also view a dashboard that highlights key issues that need to be resolved.

Image showing an area and service quality

The question the engineer now faces is which specific node should be examined to resolve issues. Watson cognitive services can take the different parameters and optimize them to find the best solution, as shown in the following image.

Alert types and options

Under normal circumstances, you would simply address the highest severity issues. With Watson, parameters such as severity, number of customers who are impacted, and revenue impact can be set to identify the key nodes that need to be addressed first. The engineer can then select these nodes and begin a dialog with Watson similar to the BGP scenario shown previously.

You have seen how users can interact with Watson in the digital agent use case by using natural language and how Watson can provide simple 1-2 sentence responses or more complex responses from a larger corpus of documents. However, Watson can also get additional information from back-end systems such as CRM systems to ensure that it provides the best response.

The network operations use case highlights the following cognitive functions of Watson:

  • The ability to integrate contextual information
  • The ability to find related concepts when the user has a dialog with Watson so that the user might think of more areas to investigate, which might not have been considered previously
  • The ability to provide multiple answers including an assessment by Watson of what the best answer might be
  • The ability to obtain feedback from users on an answer so that future responses can be improved

Personalized TV

Business problem:

In this era of digitized content, the consumer’s expectation is to consume relevant content where, when, and how they want it. This raises various issues for the content provider, the largest of which is customer retention. If the consumers feel the content provider does not understand their needs or it is difficult for them to find the content they like, they will likely look elsewhere. Therefore, it is essential that the right content be provided to the customer based on customer interest and content availability. To do this, you must understand the customer and the content you have that might be relevant for the customer.


Some of the key benefits of Personalized TV include:

  • Increased customer retention, resulting in increased revenue because customers are interested in the content that the provider can make available
  • Much better understanding of the content that the content provider owns due to the extensive analysis done on the content

Use case:

The use cases can be split into two primary areas: understanding the content and understanding the customer.

To understand the content, you must be able to perform a deep analysis of the content and identify the appropriate metadata that can be associated with the content. This metadata can then be correlated with user interest. You can use Watson to identify the video tags that are associated with the content.

Watson-identified annotations

In this instance:

  • Watson analyzes the content and displays the results. The system recognizes that we are looking at a female adult, that it is a head-and-shoulder shot, and that the person appears to be between a certain age range.
  • The system is also trained to recognize other features such as the company logo.
  • The system, out of the box, can recognize certain personalities and recognizes that this person is Lisa Kudrow.
  • The system is now able to go to the Media Asset Management Systems and other systems to obtain additional information about Lisa Kudrow.

An extra piece to understanding the content is to understand the audio. Watson can convert the speech to text and then analyze the content. Let’s assume that we would like to identify scenes in our movie where the actors appear to be angry. A conversational dialogue that occurs between the actors might be:

Image showing 4 lines spoken by 2 actors.

Watson is able to determine the tone of these dialogues by processing the information and identifying key tones.

Image showing emotion, language style, and social tendencies.

Image showing the tones of the actors

A statement such as “How can I delete this account?” by itself appears to be neutral. However, if you examine this statement in the context of the previous statements, Watson is able to determine that the tone is an angry one.

Understanding the customer includes various aspects. Much of this understanding falls into the traditional realm of analytics where data is sourced, transformed, stored, and then analyzed by various descriptive and predictive analytics tools. Watson can extend these aspects by providing various cognitive services. For example, the Watson Personality Insights service can provide a person’s psychological profile with as little as 2000 words that are written by that person — one way to get this information is with a person’s tweets. The Watson services can be combined with the traditional analytics to provide an even more comprehensive view of the customer.

Image showing someone's personality insights

Let’s briefly recap the use case and show how it benefitted from the cognitive capabilities of Watson.

  • Watson can analyze video and identify key features in the video.
  • Watson can analyze the audio and provide tone and other characteristics of the audio.
  • Watson can analyze user data beyond traditional analytics and provide psychological and other characteristics of the user to provide a more comprehensive view of the customer.
  • Watson can provide an interface to interact that uses natural language for a user to identify content that is relevant for the user after the video analytics and 360-degree view of the customer is created.


We provided a brief overview of some key use cases where cognitive computing might be used in the Telecommunication and Media & Entertainment industries. Identifying such use cases, including the business problems that are addressed and business benefits that are derived by implementing the use cases, is an important first step in the cognitive journey. It is also important that these solutions are not developed in silos, but instead use a platform where the data, components, and use cases can be used across the organization. We have not yet mentioned any of the underlying technologies behind the use cases or provided any implementation details. In the next set of tutorials in the series, we will take a deeper look into these areas.