Episode #2 - Avi Upreti, Principal, AI Strategist at Hebbia

I'm going to take a step way back and really think about how we view the investment banking business at the highest abstraction and then what we built at Hebia in order to help support what goes on.

IB ultimately can be broken down into two buckets.

This big bucket on the left side that I'll call coverage and turning that coverage into origination.

And then on the other side is this big bucket of taking that origination through to mandate execution and close.

Within these buckets, there are senior banker workflows and focus areas, and there are junior banker workflows and focus areas.

We're built to solve problems in this view of the world, right? How do you help a senior banker cover more deeply, originate more, and execute more successfully? And how do you help the junior banker who's involved in this execution focus on more higher order problems than simply doing what bankers were doing yesterday and day before yesterday? So I'll give you a couple of examples here.

Our Max product, the agent chat harness over which you can chat, is built not as a general purpose AI, but to feel more like a vertically trained partner.

So I was a fintech banker.

When a fintech banker asks about a certain company within his or her industry, they're asking about some specific things that they want to know about.

They want to know about, for in a payments company, gross merchandise volume.

They want to know about unit economics per transaction.

They want to know about contract portability, so on and so forth.

Max is built in a way such that

it knows what to look for and how to present that information.

So it starts feeling like you're talking to not an AI tool, but a coworker.

A senior banker might use that to ask about, for example, hey, what happened to the oil markets yesterday? And Max returns a result, not only in a factual way in terms of here's what happened, but also tells you about, hey, here's some downstream impacts that might impact your industry.

A senior banker might also be able to do that through email.

He might send an email to Max and say, hey, tell me about the oil market and then chat over it.

How does that impact this industrials company that I cover who has fuel price contracts and whatnot? So that's sort of how we built a senior workflow, how we view the senior workflow rather.

On the junior side of things, ultimately, the work product tends to be PowerPoints, that they tend to be Excels.

Increasingly, we're seeing more and more use of HTML dashboards for live refreshable dashboards.

We offer all of that, right? And ultimately, the goal there is to bring the junior banker going from zero to one on decks and models and so on, such that the one to 10 is spent more on really interrogating the information that is presented to the junior person rather than actually getting that information together.

In order to do that, you have to have reliable, citable information.

And we have a robust citations infrastructure so that you know that the information that is being presented to you upon which you will do the analysis and run the interrogation of that analysis is robust.

And that's how we're approaching the market.

I'm going to take a step way back and really think about how we view the investment banking business at the highest abstraction and then what we built at Hebia in order to help support what goes on.

IB ultimately can be broken down into two buckets.

This big bucket on the left side that I'll call coverage and turning that coverage into origination.

And then on the other side is this big bucket of taking that origination through to mandate execution and close.

Within these buckets, there are senior banker workflows and focus areas, and there are junior banker workflows and focus areas.

We're built to solve problems in this view of the world, right? How do you help a senior banker cover more deeply, originate more, and execute more successfully? And how do you help the junior banker who's involved in this execution focus on more higher order problems than simply doing what bankers were doing yesterday and day before yesterday? So I'll give you a couple of examples here.

Our Max product, the agent chat harness over which you can chat, is built not as a general purpose AI, but to feel more like a vertically trained partner.

So I was a fintech banker.

When a fintech banker asks about a certain company within his or her industry, they're asking about some specific things that they want to know about.

They want to know about, for in a payments company, gross merchandise volume.

They want to know about unit economics per transaction.

They want to know about contract portability, so on and so forth.

Max is built in a way such that

it knows what to look for and how to present that information.

So it starts feeling like you're talking to not an AI tool, but a coworker.

A senior banker might use that to ask about, for example, hey, what happened to the oil markets yesterday? And Max returns a result, not only in a factual way in terms of here's what happened, but also tells you about, hey, here's some downstream impacts that might impact your industry.

A senior banker might also be able to do that through email.

He might send an email to Max and say, hey, tell me about the oil market and then chat over it.

How does that impact this industrials company that I cover who has fuel price contracts and whatnot? So that's sort of how we built a senior workflow, how we view the senior workflow rather.

On the junior side of things, ultimately, the work product tends to be PowerPoints, that they tend to be Excels.

Increasingly, we're seeing more and more use of HTML dashboards for live refreshable dashboards.

We offer all of that, right? And ultimately, the goal there is to bring the junior banker going from zero to one on decks and models and so on, such that the one to 10 is spent more on really interrogating the information that is presented to the junior person rather than actually getting that information together.

In order to do that, you have to have reliable, citable information.

And we have a robust citations infrastructure so that you know that the information that is being presented to you upon which you will do the analysis and run the interrogation of that analysis is robust.

And that's how we're approaching the market.