Converging simple and complex metrics to get the most out of your conversational intelligence software
The modern digital world provides consumers with more choices than ever, meaning their expectations have never been higher. Users demand personalized experiences across devices and channels. Rather than a nice-to-have, it has become expected that companies leverage key customer experience tools, including developing methods for measuring its efficacy.
The COVID-19 pandemic has also inspired a stronger push for self-service and online support options, reinforcing the standard for around-the-clock customer service. To help fill these needs, conversational intelligence software and chatbots are being used across industries for appointment booking, medical triage, banking, and all manner of customer support interactions. Their usage growth is exponential, with one industry prediction expecting that “… consumer retail spend via chatbots worldwide will reach $142 billion—up from just $2.8 billion in 2019.”
While leveraging automated and AI-powered tools can indeed help organizations more effectively service customers, conversational intelligence software presents unique challenges around natural language and unstructured data processing. In a recent webinar, we discussed not only the best ways to design, develop, and train your semantic models and chatbots but how to better measure performance to reveal insights and opportunities that will help you fine-tune your customer’s journey.
Building conversational intelligence (or how to train your dragon chatbot)
The goal of many chatbots and conversational intelligence software as a whole is to improve the customer journey, so you must have a clear understanding of the routes your customers may take. That’s why, when training a chatbot, the best place to start is your existing data. You can dig into the top reasons your customers contact your business and leverage semantic analysis to determine customer intent and related training phrases. Start simple, and build complexity as you go.
As you design your conversational flows, you want to focus on providing an intuitive experience. Artie Merrit, Global Segment Leader for Conversational AI at AWS advises that “…the best ways to do that are to greet the user, welcome them, and let them know what the chatbot can do to try to help guide them.” For text-based systems, this can include providing quick replies and prompts.
But what do you do when the chatbot doesn’t understand the user input? How well-trained is your conversational intelligence software to understand the different ways people may ask for the same information? You need fallbacks to keep the user from hitting a dead end, but longer term, how can you use this information to improve your flows and build a smarter chatbot? This is where natural language understanding allows us to dive deeper than simply processing.
This is where context and AI-driven personalization can help your bot create more relevant experiences for customers. It also means that your various channels and customer experience tools need to share data in appropriate ways to successfully achieve consistent levels of support. One example of this would be integrating CRM data with your chatbot platform. Doing this would make it possible for the bot to pick up on whether or not a conversation has been initiated by a new or existing customer, making it easier to offer appropriate solutions.
Ask yourself: “What do you know about the user already? What are they currently doing? What have they done in the past, because this will help speed up the experience and make it much more enjoyable.” – Artie Merrit, AWS
Plan to connect your tools for data integration, but keep GDPR/CCPA compliance in mind when you do.
Measuring for success: conversational intelligence insights and opportunities
Once you have chatbots in production, you’ll want to monitor them for performance and fine-tune them. There are a number of key metrics that can be used to measure success, however more can be derived from the data itself than baseline performance. Your data, when combined, can provide deeper insights and opportunities to build seamless customer experiences.
First, let’s look at five categories of insights that you can derive from your chatbots to help you improve and refine them. These include:
How are your customers interacting with your bot?
Which bots are capturing which intents?
Do customers like your bot?
Is this task better suited for bot or website self-service?
Does your bot need enrichment?
The goal of each of these insights is to look deeper into your metrics and combine them to help you better answer each of these questions. By utilizing those answers, you can better determine if your QA bot is ready for primetime, and if your production bots need further refinement. With custom dashboards, you can pull together your key metrics and visualize your bot’s performance and the resulting customer journey and experience.
Next up, let’s explore the opportunities that monitoring your bots can offer when it comes to better understanding and ultimately improving your contact center operations.
Gayetri Bhattacharjeei, a SuccessKPI product expert and veteran of the contact center technology space, points out that, “…since you have put a machine in between, there are some automatic insights and aggregations that the machine can bring to you if it understands your customer conversations.” These insights include:
Conversation Analysis: Gain a deeper understanding of your business and customers by analyzing conversations, including the ones your chatbot is leading. These insights can include customer sentiment about your product or brand, reasons they want to cancel a service or subscription, or any trends happening within your business.
Workforce Planning: Many organizations fail to include self-service data in their workforce planning dashboards, however this can provide key insights by looking at occupancy rates, levels of self-service, and containment. Containment data can help determine effectiveness of your chatbots, assuming that they have an appropriate escalation path built in. By better understanding customer service trends that include self-service, you can more efficiently plan staffing and workforce needs, as well as identify additional tasks that chatbots may be well-positioned to handle.
Collaboration: One final opportunity is a collaboration between automated assistants and contact center agents. This would allow for agents to provide contextual assistance while also giving them the opportunity to rate a bots efficacy. This can help determine what FAQ or Knowledge Base data is presented to a customer based on context, sentiment, and conversational intelligence. This also provides you with the opportunity to utilize the same piece of AI to support both internal and external needs while getting direct feedback.
Taken together, these insights can help you build best in class chatbots, leverage your conversational intelligence software to its full potential, and foster high customer satisfaction and loyalty with your customer experience tools.
To learn more about what our experts have to say as well as how SuccessKPI can help you effectively integrate these tools in your business, sign up for the webinar replay.