The rapid development of AI technology has brought new opportunities for businesses to improve their customer service experience with Conversational AI Virtual Agents. The adoption of contact center solutions shows that businesses see the benefits, but realizing value on the investment is dependent on more than just finding the right technology stack.
1904labs’ Director of AI-Enabled Customer Service Solutions, Tom Thelen, and Solution Methodology and Research Lead, Carol Righi, recently discussed the steps that are necessary to get contact center AI efforts right the first time and realize immediate value in a webinar entitled Contact Center Transformation With AI: Why Technology Is Not Enough. Here are three of the key takeaways.
Takeaway #1: You Need to Understand the Needs of the Humans and the Business
With the developments from speech recognition and DTMF systems (think “press 1 for this, press 2 for that”) that started helping automate the customer experience in the 1960s to the AI chatbots, real-time sentiment analysis, and support tools of today, it’s now much more possible to provide a smooth experience for the customers and agents alike.
In order to do that, though, you need to think about more than just the technology and what it can do. You need to think about what it should do. At 1904labs, we talk about the three legs of the stool - the technology, the humans, and the business - that all need to be considered equally.
The technology investment needs to align to business goals, and you need to bring together key stakeholders for a measurable outcome. In the case of contact center technology, this is often related to operations costs or customer satisfaction scores.
The human aspect needs to take into account both your callers and your agents. Conversational AI Virtual Agents need to be designed so that callers will feel natural talking to them, while talking to agents and other related employees gives a view as to their wants and needs that the investment should meet.
Takeaway #2: Conversational UX Design is Necessary to Create Effective Virtual Agents
Another name for Conversational AI Virtual Agents is Intelligent Virtual Agents (or IVAs). Virtual agents aren’t intelligent by default, though. They need to be trained in order to speak to callers in an intelligent manner that meets the callers’ needs.
Conversation feels easy and natural to us humans, but that’s because we’ve been doing it nearly from birth. Conversational UX design teaches virtual agents how to converse like a human. This type of design is a relatively new field, but the principles of other human-centered fields, such as human-centered design and broader UX, largely apply. It has to start early (before development!), so you can learn about your users, the tasks they need to perform, and their wants and needs up front.
You need to engage with real, representative users and past conversations so that you aren’t just assuming how they take place. Then, once you have gone through the initial conversation design process, you should have a prototype that allows you to get feedback before it’s released more broadly, allowing you to create a style and standards by which you abide.
In short, Tom and Carol recommend contact center leaders start with these steps:
- Contact center discovery
- Dialog and conversation design
- Virtual agent personality definition
- Initial conversational UX testing
- Conversational UX design style guide creation/refinement
Takeaway #3: There Are Specific Considerations Conversational UX Designers Have to Think About
Day-to-day conversations are easy because we never have to think about them too deeply. You don’t need to think about how you start a conversation or how you react to a statement. Conversational UX designers need to think about all of this, though.
Along with just the regular quirks of conversation, there are a number of considerations specific to virtual agents that conversational UX designers have to think about. These include:
- How formal should the virtual agent be?
- Should the virtual agent ever apologize? If it does, how frequently should it apologize without it becoming too much?
- How many times can the virtual agent not understand the caller before transferring them to a live agent?
- How should the virtual agent show understanding? Should it repeat back what was said or instead show implicit confirmation by moving the conversation forward?
- Should the virtual agent tell the caller at the start of the call that they can use their keypad at a menu or to provide their phone number?
- Should the caller be able to interrupt the virtual agent mid-utterance? Does the virtual agent need to tell the caller that they can interrupt if you do allow the caller to?
The answer to these questions are often “it depends on your specific business, your industry, and the reasons for calling.” Some standards are starting to emerge, though, such as only having the Virtual Agent repeat back what a caller said for explicit confirmation at important junctures of the conversation so that the call still feels natural.