Practifi Revisited with Adrian Johnstone | E447

So you used a lot of terminology in there that I think is key that we go back to, right? So I just had a conversation with a foundry out of Poland, of all places, just before this call, and it had to do with the opportunities they're seeing.

And we harped about the BS in the AI world, right? The entire people think they can vibe code anything, even though they don't have the subject matter expertise that will get them 12 layers deep that they need to go.

So put that aside.

the entire unstructured data problem being solved.

That's really funny.

You don't know it's solved until you find out that they got solved improperly.

So you know what's better than unstructured data? Structured data.

Cannot emphasize this enough.

Knowing your data is resilient and right is fundamental to not screwing up.

And especially in our industry, when we talk about people's money, we have massive liability around that.

So we can't play games with that.

And then the last piece is deterministic versus probabilistic, right? Yes, the genie machine basically gives us all these magical answers.

But when you understand that it's basically just a lot of random calculations about probabilities.

And you can ask the exact same question in two different ways and get completely different answers because the words you use ended up resulting in a different understanding.

The reality is that's not conducive to efficiency, right? Nothing beats the efficiency of a button, right? So trying to marry the strength of the traditional with the new technology is so vital to success at the next level.

So you've touched upon a lot of those and specifically like how you thoughtfully went through all those.

And I love how you did go to, hey, we started with the old stuff and we added a bolted on the base use cases that we've seen from AI.

But then you went back to the drawing board.

So I want to go back to that entire, like going back to the drawing board.

At what point did you say to yourself, enough's enough? What we built is great, but what we need to build is completely different.

Like where was that hard? Because that's not an easy thing to do for an incoming software.

Yeah, it's really not.

And we refer to it internally as the burn the boats moment.

And we have NIA, we continue to support NIA, continue to support the additive intelligence suite.

And we did that because not every firm is at the same point on the AI readiness journey.

So you used a lot of terminology in there that I think is key that we go back to, right? So I just had a conversation with a foundry out of Poland, of all places, just before this call, and it had to do with the opportunities they're seeing.

And we harped about the BS in the AI world, right? The entire people think they can vibe code anything, even though they don't have the subject matter expertise that will get them 12 layers deep that they need to go.

So put that aside.

the entire unstructured data problem being solved.

That's really funny.

You don't know it's solved until you find out that they got solved improperly.

So you know what's better than unstructured data? Structured data.

Cannot emphasize this enough.

Knowing your data is resilient and right is fundamental to not screwing up.

And especially in our industry, when we talk about people's money, we have massive liability around that.

So we can't play games with that.

And then the last piece is deterministic versus probabilistic, right? Yes, the genie machine basically gives us all these magical answers.

But when you understand that it's basically just a lot of random calculations about probabilities.

And you can ask the exact same question in two different ways and get completely different answers because the words you use ended up resulting in a different understanding.

The reality is that's not conducive to efficiency, right? Nothing beats the efficiency of a button, right? So trying to marry the strength of the traditional with the new technology is so vital to success at the next level.

So you've touched upon a lot of those and specifically like how you thoughtfully went through all those.

And I love how you did go to, hey, we started with the old stuff and we added a bolted on the base use cases that we've seen from AI.

But then you went back to the drawing board.

So I want to go back to that entire, like going back to the drawing board.

At what point did you say to yourself, enough's enough? What we built is great, but what we need to build is completely different.

Like where was that hard? Because that's not an easy thing to do for an incoming software.

Yeah, it's really not.

And we refer to it internally as the burn the boats moment.

And we have NIA, we continue to support NIA, continue to support the additive intelligence suite.

And we did that because not every firm is at the same point on the AI readiness journey.