Planning your assistant

1. Select an initial channel

Before you decide which specific topics to build into your assistant, you must first decide where to deploy your assistant so that customers can easily find it. For example, you can embed the assistant in your company website or add it to a messaging platform such as Facebook, Slack, or WhatsApp. The primary channels that you can use to communicate with your customers are your website and the phone.

If you choose to use web chat to communicate with customers on your website, decide on which of your web pages you want the assistant to appear. To start, you might identify the pages where your customers most frequently ask questions from your customer service team.

If you have an existing interactive voice response (IVR) system with a branching structure (“press 1 for billing, press 2 for payments”), you might choose the phone integration as your initial channel. You can integrate watsonx Assistant with your existing system to automate the IVR experience so customers can talk with your assistant over the phone.

Understanding where you will deploy the assistant before you begin can help you author the right types of answers for a given channel or platform. For more information about ways to deploy an assistant, see Adding integrations.

2. Pick your assistant's domain of expertise

Decide which general domain of expertise you want your assistant to cover (for example, billing support or scheduling appointments). To make an informed decision, review any support call logs that you have access to or ask your customer service representatives. After you choose a domain, be sure that it aligns with a channel that you can control and change. For example, don’t choose to automate billing support questions if you're unable to add the web chat client to the billing web pages.

After you select a domain, you can decide which specific questions or tasks the assistant will help customers with. Start small. Pick one or a handful of customer issues that will deliver the highest value to start. It might be valuable for your assistant to answer a simple question that is asked all the time. Or maybe there's a task, such as scheduling appointments, that you can offload to the assistant to tackle incoming customer requests.

Choose a narrow set of user goals first. If you start small and choose the goals with the highest impact first, you'll have room and time to grow the expertise of your assistant. After your assistant is live, the built-in metrics of active user conversations help you understand what your customers are asking about, how well your assistant is able to meet their needs, and what to focus on next.

3. Choose the tone and language of your assistant

Before you build your assistant, it can also be helpful to choose the tone with which your assistant communicates. Is your assistant an optimist or a pessimist, a shy intellectual type, or an upbeat sidekick? Write conversations that reflect your assistant's personality. Don't overdo it by sacrificing usability for the sake of keeping your assistant in character. Strive to present a consistent tone and attitude.

Never misrepresent the assistant as being a human. If users believe that the assistant is a person, then find out it's not, they are likely to distrust it. In fact, some US states have laws that require chatbots to identify themselves as chatbots.

Decide whether and how you want to handle more than one spoken language. For more information about ways to approach language support, see Adding support for global audiences.

4. Connect to content sources

The answer to common questions might already be documented somewhere in your organization's technical information. Plan whether you want to connect the assistant to existing content sources by taking an inventory of relevant help content (for example, product information, knowledge articles, FAQs) available to your customers.

You can give your assistant access to this information by adding a search integration to your assistant. The search integration uses IBM Watson® Discovery to return smart answers to natural language questions.

5. Plan your handoff strategy

Finally, prevent customers from hitting dead ends by escalating conversations to a human agent if the assistant isn't able to resolve a customer's question or problem. As part of your planning, figure out what your escalation strategy is going to be. This strategy will vary depending on the deployment channel you choose.

If you deploy to your website, you have three options for escalating to a human agent:

• Inline
• Email address
• Phone number

If you opt for the inline approach, you can directly escalate to a human agent in your existing contact center tool without forcing the user to leave the web chat widget. This approach also provides the agent with the full context of the conversation. To use either the email address or phone number approach, provide the user with the agent’s email address or phone number. These approaches are simple to set up, but they can be more disconnected experiences for your customers.

If you deploy by using the phone integration, you can reach a human agent only by transferring the phone call to someone who can help.

Start building an assistant

If you're ready to start building an assistant, see Editing actions for more information.