After Jensen said all that crap in recent podcasts about AGI, I randomly ended up checking NVIDIA reports.. I don’t even know why.

If you’re a Jensen fan and feel like I went a bit too far there, yeah, maybe. But he’s way, way more optimistic. Like, I get it, you’re selling AI and GPUs.. but at least tell us the reality.

Now, coming back to the report, NVIDIA published its sixth annual State of AI in Financial Services report. And if you work in tech, finance, or just care about where AI is heading.. this one is worth your attention.

Because it’s backed by real survey data from 839 professionals across the financial industry: asset managers, fintech founders, investment bankers, insurance companies, regulators.. the full spectrum.

And what the data shows is simple: the financial industry isn’t experimenting with AI anymore. It’s deploying it. At scale.

65 percent of respondents said their organization is actively using AI.

Not “planning to.” Not “evaluating.” Actively using.

That’s up from 45 percent in 2024. In just one year, the number jumped by 20 percentage points.

And if you zoom into large companies, those with over 1,000 employees, that number climbs to 76 percent.

What this tells you is that AI in financial services is no longer a pilot project. It’s running in production. It’s processing documents. It’s talking to customers. It’s making trading decisions.

And here’s what makes it even more interesting.. only 11 percent of respondents said they’re not using AI and have no plans to. That means roughly 9 out of 10 financial organizations are either already using AI or actively working toward it.

So clearly.. we are not in a bubble.

This year’s report added a new section, and I think it’s the most important one: agentic AI.

For those unfamiliar, AI agents are systems that don’t just respond to prompts. They can reason, plan, and execute complex tasks on their own.

42 percent of respondents said their organizations are already using or assessing agentic AI.

Within that group, about half said agents have already been deployed. Another 41 percent said deployment would happen within the next year.

So what are they using these agents for?

The top use case was knowledge management and retrieval, cited by 56 percent. Internal process optimization came next at 52 percent. Then customer support automation at 43 percent.

But it’s not all smooth. The biggest challenge with AI agents? Performance and reliability issues, chosen by 34 percent. Followed by a lack of internal skills to manage and monitor agents at 33 percent, and data-related concerns at 30 percent.

This makes sense. Agentic AI is powerful, but it’s still in its early chapters. Companies are figuring out how to trust systems that make decisions autonomously. And that trust doesn’t come overnight.

Financial services is a perfect industry for AI if you think about it. It runs on language, numbers, documents, and patterns. That’s literally what AI is built to handle.

The survey confirms this. AI use cases are spread across the entire operation.

Customer experience and engagement was the top use case at 42 percent. Document processing was right behind at 40 percent. Then algorithmic trading at 32 percent, document management at 30 percent, and risk management at 29 percent.

The report also listed the top AI use cases that actually delivered the best return on investment.

Document processing took the number one spot for ROI at 32 percent. Customer experience was second at 30 percent. Document management came third at 23 percent.

Different segments had different priorities too. Capital markets firms pointed to algorithmic trading as a top ROI driver. Fintech companies highlighted risk management. Consumer finance companies cited fraud detection and anti-money laundering.

What this tells you is that AI isn’t just a “one-size-fits-all” tool in finance. Different types of financial institutions are finding value in different places. And that’s a sign of maturity.

Let’s talk numbers. Because at the end of the day, businesses care about ROI.

89 percent of respondents said AI has helped both increase revenue and reduce costs.

64 percent reported that AI increased their revenue by more than 5 percent. And 61 percent said it helped cut costs by the same margin.

When asked how AI improved their business, 52 percent said operational efficiency. 48 percent said employee productivity, which more than doubled from 22 percent in the previous year. And 37 percent said improved customer experience.

83 percent of respondents reported seeing a clear return on investment from their AI use cases.

So the question isn’t whether AI works in financial services anymore. The data says it does. The question is how fast companies can scale what’s already working.

This one surprised me a bit.

84 percent of respondents rated open-source software as moderately to extremely important for their AI strategy.

And nearly half of those in management roles said open source is very or extremely important.

Why?

Because enterprise-grade AI often needs models fine-tuned for very specific tasks. Generic large language models can’t always handle the nuance that a bank or an insurance company requires. Open-source models give organizations the ability to customize, own, and secure their AI without relying entirely on third-party providers.

And there’s an economic angle too. As reasoning models get more expensive with rising cost-per-token pricing, owning a model becomes cheaper long term than renting one.

Large banks and financial firms are starting to move away from purely managed AI services toward open-source foundation models for their most critical use cases.

Here’s a trend that flew under the radar but is actually huge.

47 percent of respondents said their organization uses hybrid architecture for AI workloads. That’s up from just 26 percent in 2024. Almost doubled in one year.

What does hybrid mean here? It means companies are running some AI workloads in the cloud and some on-premises. They’re mixing environments to balance cost, performance, and data security.

At the same time, cloud-only deployments dropped from 57 percent to 42 percent. On-premises-only went from 16 percent to 12 percent.

The main reason? Cost optimization. 41 percent of respondents said the top benefit of hybrid architecture is the ability to run different workloads in the most efficient environment.

This makes a lot of sense. Not every AI task needs the same infrastructure. Some tasks need the speed and flexibility of the cloud. Others need the security and control of on-premises systems. Financial institutions deal with sensitive data every day, so having the option to keep certain workloads in-house while scaling others in the cloud is a practical move.

Data-related issues were the number one challenge for AI adoption in 2025, cited by 40 percent of respondents. That’s up from 33 percent in 2024.

The specific concerns? Privacy, data sovereignty, and data scattered across different systems and locations.

The second biggest challenge was a shortage of AI experts and data scientists at 35 percent. And the third was regulatory and ethical concerns at 25 percent.

But there’s a positive signal here too. Having sufficient data for model training used to be a top concern. In 2023, 49 percent of respondents flagged it. In 2024, that dropped to 31 percent. In 2025, just 16 percent.

That tells you institutions are getting better at collecting and organizing their data. The challenge has shifted from “do we have enough data?” to “how do we manage it responsibly?”

Nearly 100 percent of respondents said their AI investment would either increase or stay the same in 2026. 83 percent said it would increase. And 44 percent said the increase would be more than 10 percent.

Where is that money going?

The top spending priority was optimizing existing AI workflows and production cycles, cited by 41 percent. That’s up from 26 percent in 2024.

This is an important detail. Companies aren’t just throwing money at new AI experiments. They’re investing in making their current AI systems work better.

The next priorities were identifying additional AI use cases at 34 percent and building access to AI infrastructure at 30 percent.

One more thing worth noting.

When asked about their top AI focus area, 68 percent of respondents said data analytics. It’s been the top answer for years, and it grew by 11 points from 2024.

Generative AI came second at 61 percent, up from 52 percent. Agentic AI debuted at 42 percent.

So while everyone talks about generative AI and agents, the workhorse of AI in financial services is still good old data analytics. Processing transactions, detecting fraud, analyzing markets, managing portfolios. The foundational stuff.

Generative AI is growing fast, though. Outside of China, its adoption in finance went from 50 percent in 2023 to 73 percent this year. That’s a massive jump in just two years.

This report isn’t just a collection of stats.

The financial industry has crossed a line. AI is no longer something companies are “exploring.” It’s something they’re budgeting for, deploying, and measuring ROI on. The fact that nearly every respondent plans to maintain or increase AI spending in 2026 tells you everything about where the confidence level is.

But two things stand out to me.

First, the agentic AI wave is just beginning. 42 percent adoption in year one is significant. If this follows the same trajectory as generative AI, we could see agentic AI become a standard part of financial operations within two to three years.

Second, the data challenge isn’t going away. It’s getting harder. As AI use cases multiply, the data problems multiply with them. Privacy, sovereignty, fragmentation. These aren’t technical problems. They’re organizational and regulatory ones. And they’ll take time to solve.

The financial industry is one of the most data-rich sectors on the planet. Whoever figures out how to use that data responsibly and efficiently will have an enormous advantage.

And based on this report.. it seems like the race is already well underway.

What do you think? Is your company already using AI in its financial operations? Or are you still in the “wait and watch” phase?