Users criticized generative AI support bots for optimizing on deflection and CSAT scores rather than accurate query resolution.
Based on 1 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@stevesi the tell is the optimization target. support bots are tuned on deflection and csat, not resolution correctness. they're rewarded for sounding final, not being right. it's not built to close the request, it's built to look closed
Product support generative AI responses are great. You get a full page answer in 5 seconds. But the answer is designed to close out the request. So then you have to ask your own LLM to figure out holes in the argument to keep this going.
The future is that your personal AI talks to the customer service AI and they hash it out for as long as it takes. https://twitter.com/stevesi/status/2076297576068878565
@stevesi the tell is the optimization target. support bots are tuned on deflection and csat, not resolution correctness. they're rewarded for sounding final, not being right. it's not built to close the request, it's built to look closed
Product support generative AI responses are great. You get a full page answer in 5 seconds. But the answer is designed to close out the request. So then you have to ask your own LLM to figure out holes in the argument to keep this going.
The future is that your personal AI talks to the customer service AI and they hash it out for as long as it takes. https://twitter.com/stevesi/status/2076297576068878565
Users criticized generative AI support bots for optimizing on deflection and CSAT scores rather than accurate query resolution.
Based on 1 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.