The entire AI customer support industry has pretty much already converged into this model. Tickets closed without being escalated to humans are usually what’s defined as an outcome. We think this model makes the most sense for any vertical AI company where agents are actually completing tasks end to end.
Success is defined by the buyer and seller before they use us. We just facilitate the parameters they agreed upon, so it’s pretty variable. A successful outcome can be anything from sourcing a real estate property that ends up closing to finding $X in cost savings for a dental clinic, using research agents. The biggest difference is you’re not just charging for tokens, but assigning a dollar figure to what a job well done looks like, no matter how many tokens it takes to get there.
>Tickets closed without being escalated to humans are usually what’s defined as an outcome.
So I can make bank with a bot that replies to every query with "** you! Ticket closed." ?
>cost savings for a dental clinic
How can you software possibly know that?
Not quite, because the buyer and seller would never agree to that being a real outcome. It would have to meet certain requirements/thresholds for customer satisfaction. That’s where we come in to verify that those contract terms are actually met as it happens.
There are a ton of ways to calculate cost savings depending on what it’s built for. Voice agents that handle booking and outbound calls, supplier sourcing agents that only buy the best priced items, etc. It’s then possible to translate those things into a monetary figure, relative to the business.
I can see why people buying/using software would want this, but why would any software startup want to use it? What are the markets (other than non-profit), where pay vs success (and not attention/friction) is the limiting factor? The idea is you’re unlocking new markets or models, right?
And when/how is success measured? Otherwise it’s just like API billing, right?