Expanding the “Suggested Intents” beta to include misunderstood messages

Our announcements at Google Cloud Next 2018 last July included new beta status for the “Suggested Intents” feature in Chatbase Virtual Analytics. This feature, which is unique in its product category in taking advantage of Google’s machine-learning capabilities, suggests intents for “not handled” messages in which none was initially detected by the virtual agent (aka missed messages). Developers can then use them to more easily fill the functionality gaps as reflected in the list of missed messages shown in the “Not handled” tab of the Messages report.


Now, we’ve expanded this feature to support misunderstood messages surfaced in the report’s “Handled” tab. For such messages, the virtual agent detected an incorrect intent, leading it to answer a slightly different request than the one asked (such as “insurance options?” triggering an intent instead of ) -- which is quite a poor user experience. For those cases, Chatbase now suggests an intent that it predicts would have let the virtual agent resolve the request correctly. As with suggested intents for missed requests, this information makes it much easier to optimize the virtual agent over time, leading to a better user experience and hence better retention.


For more background about unsupported, misunderstood, and missed requests (aka “UMM” errors), see the docs.

Joining the beta

Suggested Intents is most useful for virtual agents that cross a certain threshold for activity and complexity. If your agent speaks English and has at least 2,000 user messages per month and 20 unique variations per intent, you can request access to the beta by sending us an email. (Be sure to include your API key or bot name.) You will initially be placed on a wait list and we will activate the feature as soon as possible.

We look forward to bringing you news about the next milestone for this beta when available!

About Chatbase
Chatbase (chatbase.com) is a cloud service for more easily building, analyzing, and optimizing AI-powered customer service virtual agents. Powered by Google's machine learning and search capabilities, the Chatbase approach is faster, more accurate, and more comprehensive than manual methods, helping contact centers deploy a virtual agent that improves customer satisfaction in a fraction of the time. Chatbase is brought to you by Area 120, an incubator operated by Google.

Comments