Insights: see what your AI gets wrong, and what to fix

Roger Junior
Co-Founder, Product and Engineering

Published

2 min read
Product Update
An insight in Hay.chat: shoppers asking about gift wrapping, with the likely cause and a button to suggest a knowledge fix.

Reading every conversation to find out where your AI struggles does not scale past the first week. Insights does that reading for you. Each night Hay.chat goes through your conversations and your team's ratings and groups them into patterns: what the AI gets wrong, what works, and what to change. It is available on every plan, hosted and self-hosted.

Patterns, with the evidence#

An insight is one recurring situation, such as shoppers asking about gift wrapping and not getting an answer. It shows how many conversations it covers, how many of them ended up with your team, when it was first and last seen, and a likely cause. Under it are short excerpts from the conversations themselves, with personal data removed, so you can judge it for yourself. Positive patterns show up too.

From pattern to fix#

Each insight comes with a suggested action. Hay.chat can draft a change to your knowledge or an edit to a playbook for you to review, or write a task for your own tracker. Nothing is changed for you: a knowledge fix is shown before it is saved, and a playbook edit stays a draft until you publish it.

See whether the fix worked#

When you mark an insight as addressed, Hay.chat keeps watching. It compares the weeks after with the 30 days before and tells you whether the pattern is fixed, improved or still there. If it comes back later, the insight reopens.

Owners and admins can also get a weekly digest by email.

Insights needs conversations to work with. A pattern appears after the nightly analysis, once it shows up in enough of them.

Start a 30-day free trial to try it on your own conversations.

About the Author

Roger Junior

Co-Founder, Product and Engineering

Roger designs and builds Hay.chat's architecture. Before this he was a UX/UI designer working with international development teams at global companies, and he's now the technical founder behind the product.