Brian O’Malley

You only have one major customer summit a year.

People are still at the point where they want a human in the loop in order to validate that the AI is doing the right thing because hallucinations are still a real element.

And you can prompt engineer, you can have the system double check itself.

But the intrinsic value of still having a human

5% in the loop ultimately makes a lot more people comfortable spending real dollars on these platforms.

So another way, having AI do 95% of the work is also extremely valuable.

And instead of taking 20 units of energy, you're doing one unit of energy.

So you're basically bringing it down by 95%.

Yeah, exactly.

You want to feel like there's a human helping you on the other side.

And so it's really important to have people pick up that phone initially.

But when it comes to pulling medical records, when it comes to filing paperwork with the local court system, a lot of that can be automated behind the scenes.

And even subsequent conversations, those can be pushed to chat where it's a lot easier to use AI systems than AI voice, which is still a little bit clunky right now.

There's a belief out there that AI and these large LM models will take the route of autonomous and self-driving, and the last 1% is going to take 5, 10, 20 years to perfect.

Do you subscribe to that?

And if so, why do you think that is?

I think it depends on what you're asking it to do.

There are certain tasks with AI where 95% right is really 100% right.

Like you think about writing a blog post or writing an email to someone.

It might not be the exact words, but it's close enough that it solves the goal of what you're trying to accomplish.