High performing customer support chatbots are a massive unlock thanks to LLMs

Previously, chatbots were clunky, and no fun to talk to

Published: July 18th 2025

Previously, chatbots were clunky, and no fun to talk to

LLMs now understand the meaning behind our words

This opens up a mass of possibilities, chatbots being one of the first that come to mind

Looking at some case studies, the numbers for customer support chatbots look good

→ 83% ticket resolving percentages at Gamma
→ 50% ticket resolution rate at Consensys, growing to 70% over time
→ $163K annual savings at Peddle by reducing support volume by half

I don't find numbers surprising

Chatbots sit right at the natural language to computer interface, which is exactly what LLMs have unlocked for us

Things that were previously not possible, are now very doable

Some of the pricing structures around these chatbots are interesting also. For example, Fin, the support bot from the well known messaging app Intercom, charges on a resolution basis

→ $0.99 per ticket resolution
→ of a minimum of 50 resolutions per month
→ pay for what you get, basically

Looking at the case studies, it does takes time to set them up

→ You need a few months to train and improve the chatbot so it can answer questions well.
→ And even after that, your team needs to keep it updated so that it can handle new questions.

So to get the ROI on the initial investment and ongoing costs, you will definitely need some threshold of customer support tickets for it to make sense.

If your business does get a lot of customer support questions every month, I think it would definitely be worth exploring if these AI powered chatbots can help streamline your customer support.