By Ofer Ronen, Chatbase GM
Around six months ago, Chatbase became generally available. In our journey thus far, we’ve processed hundreds of thousands of bots, and billions of messages and intents. Having this data provides a unique window into emerging trends in the rapidly growing conversational analytics industry.
To mark this milestone, we want to share some observations about why people are building conversational interfaces (aka bots), and why some bots are more successful than others.
One can score bot use cases based on where they should lie across a spectrum of user expectations about engagement. Most value is found at the edges:
Around six months ago, Chatbase became generally available. In our journey thus far, we’ve processed hundreds of thousands of bots, and billions of messages and intents. Having this data provides a unique window into emerging trends in the rapidly growing conversational analytics industry.
To mark this milestone, we want to share some observations about why people are building conversational interfaces (aka bots), and why some bots are more successful than others.
Virtual agents are finding a place in contact centers
As the data in the following chart (based on a sample of more than 700 bots in Chatbase that have provided a use case) suggests, bots are becoming particularly popular for customer service and support. This data point mirrors what we see in recent industry research about the growing prominence of virtual agents for customer service. For example:- Gartner has predicted that by 2020, 25% of customer support interactions will involve virtual agents or bots.
- Juniper Research predicts that AI-powered virtual agents will save banking and healthcare contact centers nearly $8 billion by 2022.
- In a new Customer Contact Week (CCW) market study, 57% of respondent organizations said they will invest in bots for “improved self service” in 2018.
The range of user engagement is wide
Our data (the data in this section is based on bots with at least 10,000 messages in 2018) also suggests a broad spectrum of engagement “intensity” across bots in every industry/use case, whether measured by number of turns (median: 2) or session time (median: 24 seconds). Some bots boast at least 14 turns and/or 9 minutes per session, an amazing level of engagement! But as we’ll see below, intense engagement is not automatically a positive.Sometimes, less user engagement is better
This observation leads to what is perhaps the most valuable insight: More user engagement (measured by number of turns or session time) is often assumed to be a good thing. But in fact, for many uses cases, less is more -- for customer service, for example.One can score bot use cases based on where they should lie across a spectrum of user expectations about engagement. Most value is found at the edges: