Customer care use case
With Watson™ Assistant on Cloud Pak for Data, you can implement a customer care use case with conversational AI to deliver positive and meaningful experiences for customers.
When customers need support, they expect fast, consistent, and accurate answers across applications, devices, or channels, including web chats, social-media messages, text messages, phone calls, or custom applications. Every customer support interaction and experience can influence your organization's ability to acquire and retain customers. With a customer care use case with conversational AI on Cloud Pak for Data, you can create an intelligent assistant to deliver positive customer care experiences.
This video provides a visual method as an alternative to following the written steps in this documentation.
Challenges
Many enterprises face the following challenges with their customer care processes:
- Understanding what customers are saying
- Your organization needs to understand nuanced phrasing in voice and chat interactions to discover what your customers want and need.
- Delivering accurate information
- Your organization needs to deliver accurate information across channels as seamlessly as possible.
- Learning from interactions
- Your organization needs to learn from customer interactions so that you can improve responses and enhance the customer experience.
Example: Golden Bank's challenges
Golden Bank serves many customers and strives to provide the best support experience possible, which directly impacts customer satisfaction and retention. Follow the story of Golden Bank as subject experts create an online banking assistant that uses conversational AI to understand their customers' needs and deliver accurate and helpful solutions.
Process
The Watson Assistant service in Cloud Pak for Data, including optional integrations with Watson Discovery and Watson Speech services, provides the tools and processes that your organization needs to implement a customer care solution with conversational AI.
1. Identify your customers' goals and input
To set up a conversational AI solution, the first step is for your subject experts to identify your customers' goals and the questions your customers want to ask the assistant. For example, customers might ask the location of a bank branch that can handle a specific transaction. In Watson Assistant, your subject experts match the customers' goals to input content and identify the keywords and phrases. The assistant that your subject experts build, recognizes the content and variables that are mentioned in the customers' input and chooses appropriate responses and actions.
| What you can use | What you can do | Best to use when |
|---|---|---|
| Customer input
|
Plan and categorize customers' goals or purposes based on the conversational AI solution
that you are building. Input phrases represent the questions that customers ask or requests that customers make when they use the assistant that you build in Watson Assistant. Add your own questions and requests and import content from CSV files. Customer input is stored as variables that are saved to work through flows more quickly, or for use across conversation topics. |
You have determined your customers' goals and potential variables and want to convert those goals to real-world phrases and questions that customers might use in their conversations with the assistant. |
| Responses
|
Identify the text or speech that the assistant delivers to a customer at a particular step
in the conversation. Depending on the step, you can add a complete answer to a customer's question or specify a follow-up question. |
You are ready to indicate the assistant's responses based on the customer's input and the conditions. |
Example: Golden Bank's customer goals
Analysts provide details on what the Golden Bank customers want to know. For example, when their customers sign in to do online banking, they might want to ask what to do when they lose a card, how to pay a bill online, where to find the closest ATM, or how to apply for a mortgage.
Golden Bank's subject experts enter real-world phrases that are related to these goals in Watson Assistant. They upload existing data from CSV files with variable data based on previous conversations with customers.
2. Design, build, and test an assistant
The next step in the process is to design, build, and test an assistant. An assistant is the implementation of a conversation flow that starts with a message from a customer that is then sent down the appropriate resolution path. Your subject experts build an assistant to use the customer inputs and assistant responses that they defined in the first step. They create actions to interact with customers and deliver the responses. As your subject experts build your assistant, they can test it out and make adjustments.
| What you can use | What you can do | Best to use when |
|---|---|---|
| Actions feature
|
Use the Actions feature in Watson Assistant to create conversations and indicate the conditions that trigger assistant response. Each action represents a task to complete or a question to answer and contains the expected the customer input and assistant responses from step 1. Create Actions from scratch or choose templates from a catalog of actions. |
You have identified your customer's input and goals and have planned out conversations and you need to specify the action flow of the assistant based on that content. |
| Preview feature
|
Preview the flow and conversation. Type phrases that are variations of your expected customer input to see whether they are recognized. Customize the chat look and feel. |
You are building or modifying an assistant and you want to see how it will work and make adjustments before customers start using it. |
| Watson
Discovery
|
Use an optional search integration with Watson Discovery that includes existing FAQs or other curated content that you own to find relevant answers to customer questions. Train AI for deep understanding of your content, including tables and images to help you find business-value hidden in your enterprise. |
You want to give your assistant access to existing data collections that it can search for answers. |
| Watson Speech
services
|
Use an option integration with Watson Speech services to connect the customer with someone from your support team through a web chat. Use speech processing capabilities to transcribe spoken text from audio or to produce natural-sounding voice audio from written text. |
You want to enable customers to get more personalized help, or to discuss a sensitive subject. |
| Voice Gateway
|
Use an optional integration with Voice Gateway with Watson Assistant to connect the customer with someone from your support team through a phone call. |
You want to enable customers to receive personalized support over the phone. |
Example: Golden Bank's assistant flow
Golden Bank subject experts add dialog nodes to the assistant to indicate the conditions that trigger actions.
When a customer uses the online assistant and types something that includes or is close to one of the conditions, a certain action is triggered. For example, if a customer types, "Where is the closest ATM?", the assistant might ask what types of services the customer requires. Then, based on the response, the assistant would provide a list of ATM locations that offer that service.
Before Golden Bank subject experts deploy their new or updated online banking assistant, they use the Preview feature to test the phrasing in the conversation and the flow of the actions.
3. Analyze and improve
The final step in the process is to analyze and improve the assistant. Your subject experts can view the history and metrics of the assistant after it is implemented to analyze the conversations and usage to determine ways to improve the experience. They can go back to steps 1 and 2 to implement changes.
| What you can use | What you can do | Best to use when |
|---|---|---|
| Analyze feature
|
View the history of conversations that your assistant has with customers and use these conversations to improve your conversational AI flow. Expand individual conversations to see what a customer said and how your assistant answered. Go through actual customer conversations and make corrections to input phrases and responses that are defined incorrectly. View the metrics overview to see the amount of traffic for a period of time and the content that was recognized most often in customer conversations. |
You have a working assistant in production and you want to evaluate its accuracy and effectiveness and make improvements to support the customer experience. |
Example: Golden Bank’s testing and improvements
After the Golden Bank assistant is built, the subject experts can use the Analyze feature to go back and adjust the wording, conditions, and actions as necessary. They can use the metrics to answer questions such as "Which phrases appeared most often last week?" and "Which variable values were recognized the most times during February?". From the history of live conversations, the subject experts can upload real data to improve the customer experience.