Using an automated question and answer (Q&A) system comes with both hype and headaches. The allure of deflecting incoming customer inquiries is enough to consider using artificial intelligence (AI) to answer incoming requests in real time. AI-powered Q&A systems come with their own set of considerations that you should know if you are thinking about deploying this kind of technology.

Know Your Why And Your Data Sources

Before starting to evaluate AI vendors, spend resources towards getting a better idea of what it is that you want to automate and why it is you want to do so. Specifically, you should start by looking for Tier 1 conversations to automate. These are the types of questions that are highly repetitive, require little back and forth between an agent and the customer and hopefully occur at medium to high volumes to address overwhelmed human agents. Examples of Tier 1 tickets include order status inquiries, reset password requests and general FAQ-related questions. What you want to understand is — how often do these Tier 1 tickets occur out of your total ticket volume and will automating them make an impact on your customer service bottom line? This will provide a goal for your AI to achieve.

Once you have a good handle on the types of tickets you wish to answer, ascertain internal (and sometimes external) sources of information that can be utilized to answer these questions. For most companies, this will usually fall into obtaining historical chat conversations to train the AI, public and private knowledge bases — FAQs, help articles, manuals, etc. — and integrations with real-time streaming information — order delivery status systems, live data-lookup application programming interfaces (APIs), etc. All of these will arm your AI with the information needed to answer your customers’ questions.

Optimizing For Customer Experience

After you have a good idea of the types of requests you wish to handle, it’s time to evaluate the types of AI that can be utilized. This highly depends on the expectations of your customers and the type of user experience you want to present to your customers. The main thing to consider is: Do you want your AI to simply surface help articles to empower your customers by giving them the tools they need to help themselves, or would you prefer to create an AI to guide them step by step and possibly even perform some (very simple) tasks for them? The latter is harder to create.

There are a couple of different ways that an AI-powered Q&A system can answer the same question. For example, if a customer has a pricing question, one AI may answer the question by sending them to a help article that will hopefully answer their question. Another AI may take the same question and engage the customer by having a dialogue-based conversation, asking follow-up questions to learn more about their question and offering to look up the information for them. It’s all about what experience you want to deliver.