I've been creating LLM based apps for past couple of years. One of the FAQ from people who are hearing this first (without seeing the app) is 'Wouldn't ChatGPT be able to do the same thing? Why should users come to your app?'. This is my thought process and a framework that I thought could be used to measure what makes the App stand out from a regular ChatGPT output. It could be the personal knowledge / context the creators have used, the experience of the app, etc. Early stage thinking, open to feedback.

I don't think accuracy differential could sustain long term. Companies are already entering health, legal verticals. Token efficiency is interesting, LLM companies do not have incentive to reduce token usage.