As someone who is also using Chatbase, I am excited to see the common discussions list; it is clear and practical. It will be a good avenue to learn and also share my knowledge with others in this forum.

You contrast problem-solving with impressive demos but you left out a third category: problems that AI creates while solving others. An AI tool that accurately deflects 70% of tickets but alienates 15% of customers may have a negative value on the business.

To answer your question, I think the most valuable insight for us has been tracking emerging topics. Customers suddenly started enquiring about a new policy change we hadn’t explicitly trained the AI bot on. Because we were monitoring topic trends rather than just resolution rates, we were able to update the AI’s knowledge base before it becomes a CSAT problem.

For us, knowledge base accuracy is the top priority for us. An AI sounding polite and confident while giving the customer the wrong information is worse than having no chat support at all.

This is a very insightful review. I love how the article explains the strengths and weaknesses of each Decagon alternative, making it easy for businesses to choose the option best suited to their specific needs.

In your closing criteria, you mentioned “produces measurable value”, but this where a lot of AI evaluations break down. How can you measure the value of work that no longer requires human efforts? Traditional SaaS metrics like usage, seats, and features adopted, do not capture it. It will be difficult to be advocating for a new evaluation framework without providing actual metrics to operationalize it.

Your point about Intercom Fin’s pricing is painfully true. We used it, thinking it would save us time, and our monthly bill was really shocking. $0.99 per resolution looks cheap until you realize the number of 2-message chats your team receives daily.