Generative AI's applications

I've always thought of myself as a creator of some sort. I was a demon on Adobe Cloud (PS, AE, etc.) and dropshipping keycaps in 7th grade. The world needs creativity and always will. It's the sole source of fulfillment for many people.

The startup space is headed for more change this decade than ever before. I've been on a lot of calls recently, and we keep coming back to the same feeling: this doesn't feel like the same world in terms of technology. Many founders aren't unreachable anymore, and there's generally so much more success than before. Not every successful company is a myth or some insane unicorn story. People are local in a digital sense, accessible, and here.

On one of those calls, someone mentioned how GenAI is already changing everything, and by the time I graduate, the world I'm walking into will probably look nothing like the one people are describing right now. Everybody makes these vague statements and we all nod along, but I came to the realization that the entire economy will actually change. We're all aware of it. The difference is I want to start understanding how that change takes place. I personally think the idea of what makes someone valuable might shift from how hard they work to how well they think. Whatever space I end up in, I'll have to rebuild my expectations from scratch, but that's fine because my expectations are constantly changing anyway.

I'm going to write about GenAI in finance specifically because finance is an industry where technology has been deeply integrated for decades. It feels like the first real technological shift touching both the analytical and creative sides of the industry at once.

The space right now

AI's presence is pretty clear in every industry that remotely involves technology. It's replacing everything we're doing. Finance is no exception.

The last year made that obvious. Startups like Rogo are building what they call "the AI analyst," which really means automating most of what a junior banker or research associate does: reading filings, summarizing earnings calls, writing company profiles, building slides. Their models run through filings and research notes and turn them into structured output ready to use. The product pitch is that it saves time, but the real change is that it redefines what an analyst even does. Rogo started with a $7 million seed round, then raised an $18.5 million Series A in 2024 and a $50 million Series B in 2025 led by Thrive Capital, bringing total funding to around $75 million. Their team is made up of former bankers and engineers who actually understand the day-to-day workflow, and they're used by large institutions handling trillions in annual deal volume. Clients include names like Nomura, Moelis, and Tiger Global.

Then there's Hebbia. They focus on what they call document intelligence, which basically means making sense of PDFs, contracts, and transcripts faster than any human could. Same technology underneath, but built for people in compliance or legal who spend their entire day looking for one line in a hundred pages. Even at trillion-dollar banks, you'll still find analysts manually checking footnotes and referencing Excel sheets line by line. Hebbia fixes that. You upload something, it cross-checks everything, and you're done. They raised about $130 million in 2024 at a valuation near $700 million, after an earlier $30 million Series A. Clients include firms like KKR, Centerview, and Permira.

Klarity is another one. They automate contract review and accounting workflows for accounting firms and corporates that go through thousands of agreements a year. It's not exciting work, but it's necessary and it takes up thousands of hours (accounting for you). Klarity's model summarizes documents, flags exceptions, and routes them automatically. They raised roughly $70 million in mid-2024.

Then you have companies like Cohere and Anthropic. They aren't building "AI for finance," but they're the foundation of the entire movement. Cohere has raised hundreds of millions to build large language models designed for enterprise use. Anthropic works with every major financial institution that wants private, in-house deployments of generative models. Every bank, asset manager, and PE firm is building internal copilots on top of these models and training them on proprietary data so insights never leave their systems. The incumbents are moving in the same direction. JPMorgan has IndexGPT and its own LLM suite. They're not buying AI tools anymore because they have the resources to build them internally. They know what's coming.

All of these products fall under "GenAI for finance," but they're solving very different problems. Some want to make research faster, some want compliance to be more efficient, some want to automate reasoning entirely. Most aren't there yet, and the tools still hallucinate and miss context. But what used to take five people now takes one, and that says a lot.

The impact they are having

I don't think people understand how fast this is actually moving. When you read about it online, it's the same vague hype on LinkedIn. The real numbers are hard to come across without digging. Rogo's platform is already being used by thousands of analysts across investment banks like Moelis and Nomura, and they say it saves roughly 10 hours a week per analyst. That doesn't sound crazy at first, but if you have 5,000 bankers using it, that's about 50,000 hours of labor saved every week. Assuming $80 an hour in cost, which is conservative, that's four million dollars in time saved every week. Over a year, that's over $200 million of productivity.

Hebbia's numbers are just as striking. Some firms report automating 70 to 90 percent of their contract review process. The average hold time for deal memos used to be days and now it's hours. Hebbia claims 92 percent accuracy on analysis tasks that humans used to get wrong 30 percent of the time. That changes the entire trust equation in research. You can rely on speed without sacrificing correctness, and that's never been the case before.

These companies aren't even fully built out yet. Almost all of them are still in a growth phase, still adding features, still improving models, and firms are already dependent on them. Analysts in every financial field will start becoming operators of AI rather than analysts in the traditional sense. That's a big shift because the whole analyst pipeline used to be built around manual repetition. You learned by doing grunt work, and now that work is absorbed by technology. The entire structure of training in finance will have to change in response.

The impact shows up in how firms operate, too. Some have quietly cut analyst headcount by 10 to 20 percent since 2023, and output has stayed flat because AI picked up the slack. The firms adopting this tech early are already pulling ahead. JPMorgan has an AI summarization tool for research reports that 10,000 employees use internally. Morgan Stanley has a similar system powered by OpenAI. When you can prepare a client deck in hours instead of days, or check an entire credit agreement in one afternoon, you move faster than your competition. Speed kills.

The total market for AI in finance is already around $38.4 billion according to Google, and they project it could reach about $190.3 billion by 2030. That's a CAGR of about 30.6 percent over six years.

Some predictions and thoughts

The biggest thing nobody is thinking about with GenAI in finance is how physical it actually is. It feels digital, but it's powered by something real. The data centers that make all this possible are huge, and they're multiplying. When Goldman or KKR uses Rogo or Hebbia, servers are being hit somewhere. Those servers sit in massive buildings, eating power and pushing heat. Every model query is energy. Multiply that by the thousands of bankers and analysts running prompts all day and you start to see how finance's "digital efficiency" might actually be raising the world's energy bill. Data centers already account for around 2 percent of global energy use, and that number could easily double in a few years if GenAI becomes standard in financial operations. The industry that prides itself on efficiency might be making the world less efficient. I don't know that they care. That's more of an existential debate than anything else.

There's also no way this technology doesn't affect jobs. The whole premise of these tools is that they replace labor. You can't make everyone twice as productive and not expect some redundancy. Bankers won't disappear overnight, but their roles will change, and there will be fewer of them. If one analyst can handle two or three times more output, firms can do the same work with fewer people. Instead of hiring for grind, they'll hire for judgment. The analyst of the future will be someone who understands data context, logic, and narrative, spending less time collecting numbers and more time interpreting what the models produce.

One thing that isn't mentioned much is how this could mess with market behavior. When everyone starts using similar AI tools, the analyses on companies might converge. You could end up with hundreds of firms drawing the same conclusions about the same companies at the same time, which would make markets more volatile because everyone reacts in sync. Imagine all the analysts reading from the same machine-summarized sentiment data. There would be too much alignment.

If this technology keeps compressing time, decisions will happen faster than people can process their consequences. Speed is generally good in finance, but markets might start moving so quickly that even regulation can't keep up. AI is making its own waves in regulation and compliance, though. Humans will figure it out.