The Secret to Scaling AI in Financial Services with Ankur Patel
Product-market fit often sounds more binary than it really is, and founders often talk about it as a moment of arrival. You search, you validate, customers react, and eventually the market tells you that you have something worth building, but in practice, the signals are rarely that clean.
People can be interested without being ready to buy, they can agree to a proof of concept without having a clear path to production, they can praise the product without becoming the kind of customer who helps the company grow.
That distinction was at the center of Collin Stewart‘s conversation with Ankur Patel, founder and CEO of Multimodal, on the Predictable Revenue Podcast.
Multimodal builds AI for document-heavy, decision-heavy workflows in financial services, and Ankur’s story is useful because it shows how easy it is to mistake activity for traction.
Interest Is Not the Same as Intent
In the early days of Multimodal, there was no shortage of interest.
Ankur had spent more than 15 years in financial services, including time at institutions such as JPMorgan and Bridgewater. When large language models began to demonstrate new reasoning capabilities, he saw a clear opportunity: complex financial workflows could be redesigned around AI, especially those involving large document sets and human-level decisions.
That insight gave Multimodal a broad surface area. Lending, account opening, compliance, fraud, underwriting, claims processing, and other regulated workflows all seemed like possible markets.
- The product could handle complex documents.
- The market was curious about AI.
- Large enterprises were willing to take calls and explore pilots.
But curiosity is a dangerous signal.
Ankur described a period where buyers across financial services and insurance showed strong interest. Some wanted proof of concepts, some wanted to understand the technology, and some were excited by the idea of AI agents before the market had really adopted that language. From the outside, that level of engagement can look like demand.
But a founder has to ask a harder question:
Is this buyer curious, or are they ready to change how their business operates?
That difference matters because early-stage companies don’t just need conversations, they need conversion, and they need a segment where the pain is urgent enough, the timing is right enough, and the buyer has enough internal momentum to move from exploration to paid use.
Customer Interviews Only Work When You Listen for Urgency
Multimodal did not start by guessing its wedge. The team interviewed buyers across financial institutions and looked for workflows that followed a specific pattern: the process had to be recurring, occur at high volume, involve complex document sets, and require a level of reasoning that had historically been difficult to automate.
Financial services is full of workflows like that, but not every painful workflow is the right starting point. In mid-2023, Ankur said the team had spoken with roughly 50 buyers, and lending showed up in about 30 to 40 percent of those conversations. That was not a perfect, overwhelming signal, but it was enough to stand out from the other options.
Banks and credit unions care deeply about deposits, lending volume, and the quality of lending decisions, so if AI could help shorten the time to quote, it could do more than reduce cost. It could help the institution win more business.
That is the kind of signal founders should pay attention to.
A workflow can be inefficient and still not be urgent, a buyer can acknowledge pain and still not have a budget, and a team can like a demo without having a reason to act now. The strongest early markets usually have more than pain; they have pressure.
The First Real Signal Is a Customer Willing to Keep Paying
One of Multimodal’s clearest early signals came from a customer working on mortgage documents. The documents were messy: pay stubs, bank statements, tax returns, varied formats, poor-quality inputs, and information that was difficult to extract reliably. The CEO was skeptical because he had tried to solve the problem before, but he was also willing to take a chance.
The ask was simple: prove it in 30 days.
That kind of moment is useful because it moves the conversation out of theory. The customer was not just saying, “This is interesting.” He was putting a real workflow in front of the team and creating a path to continue if Multimodal could solve it.
When the proof of concept worked, the important signal was not a surprise. It was willingness to move into a paid subscription, which is where many founders need to be more disciplined. Positive feedback feels good, especially when the product is early, but a successful demo or a buyer saying they have never seen anything like it is not the same as product-market fit.
The stronger signal is behavior:
Will they pay, keep using it, expand usage, bring the product into another workflow, or introduce it to another team? Ankur pointed to continued usage as one of the clearest signs that product-market fit is real, because customers are not just buying once. They are coming back and buying more.
Broad Demand Can Dilute the Company
The hardest part of early traction is that it rarely arrives neatly. Once Multimodal had proof that its product could solve complex document workflows, the temptation was obvious: why stop at lending?
The same core capability could apply to account opening, compliance, fraud detection, insurance underwriting, and claims processing. Large companies were interested, buyers wanted to explore, and the market was moving quickly, so for an ambitious founder, it is easy to interpret that as permission to expand.
But Ankur was direct about what happened:
The company got diluted in its effort. The problem was not that the other markets were fake, since some of them may become meaningful opportunities later. The problem was that not every opportunity was equally ready. Some buyers were interested in AI but not ready to move from POC to paid subscription, some teams were still learning what agentic AI could mean inside their organization, and some were taking calls because the category was exciting, not because they had a clear buying path.
That is one of the more subtle false signals of product-market fit.
A broad market can make the product feel more validated while making the company less focused, which means founders do not only have to decide whether a market has pain. They have to decide whether that market is the right place to concentrate the whole company right now.
Focus Makes the Whole Company Easier to Operate
When Multimodal refocused on the areas where conversion was strongest, the business got simpler. Ankur described how that focus clarified messaging across social, the website, outreach, conferences, and sales conversations, while also helping the product team go deeper into the workflows that mattered most to credit unions and lending teams.
That is one of the underrated benefits of focus:
It is not just a strategy decision, it is an operating decision. When the target market is too broad, every function bears the burden of complexity. Messaging has to cover too many use cases, sales has to speak to too many buyers, product has to support too many workflows, and marketing has to explain too many versions of the value proposition. Even conference strategy becomes harder because the company is unsure which rooms matter most.
Focus does not mean the company has to abandon its ambition.
Ankur was clear that Multimodal can expand into other segments and workflows later, but the immediate priority is to win the wedge before expanding beyond it. That is a practical founder lesson: expansion is easier when the company has earned it. Before that, breadth can look strategic while quietly slowing everything down.
In Trust-Heavy Markets, the Non-Scalable Thing May Be the Right Thing
For Multimodal, go-to-market fit also meant accepting that the best early channels were not always the most scalable ones. Outbound email had become less effective than it was three years earlier. At the same time, the stronger motions were more relationship-driven: channel partners, advisory boards, podcasts, quarterly reports, research institutions, and peer networks.
The advisory board lesson was especially sharp.
Ankur said he wished they had doubled down on advisory boards earlier, even though they did not feel scalable at first. That matters because Multimodal sells into a trust-heavy market. Credit unions and financial institutions are not adopting AI in the same way a Silicon Valley startup might; they operate in regulated environments where teams have to manage risk, internal skepticism, job displacement concerns, data sensitivity, and a slower adoption cycle by design.
In that environment, the right early motion may look inefficient from the outside.
Advisory boards gave Multimodal a way to bring business operators and technologists together in a peer-driven setting, where credit union leaders could talk openly about what they were trying, where they were stuck, and what responsible AI adoption needed to look like. Multimodal could participate as the technology expert, but the value was not only a vendor pitch. It was a trust-building forum.
For founders, the takeaway is simple: scalable channels are not always available at the beginning. Sometimes, the non-scalable motion is what earns the company the right to build a scalable one later.
Product-Market Fit Has to Be Defended
One of Ankur’s most useful points is that product-market fit is not permanent, especially in AI. What felt remarkable in 2023 can feel ordinary later because buyer expectations keep moving. A customer who once thought an AI workflow was impossible may now compare it against ChatGPT, Claude, Microsoft Copilot, or a wave of new AI vendors.
That does not mean the startup is doomed.
It means the company has to keep earning its position. For Multimodal, that means staying ahead on product and on go-to-market. The product has to remain meaningfully better for regulated, document-heavy workflows, but the company also has to explain that difference in a way buyers understand.
In financial services, being technically ahead is not enough if the buyer does not trust the vendor, does not understand the safe adoption path, or does not give the company a chance to demonstrate the product. That is why go-to-market fit becomes part of protecting product-market fit: the product can create the aha moment, but the go-to-market motion creates the opportunity for that moment to happen with the right buyer, at the right time, in the right context.
Go-To-Market Fit Is What Makes Traction Repeatable
The deeper lesson from Multimodal’s story is that founders should be careful not to treat product-market fit as the finish line. It is a milestone, but not the whole system.
A company can have a valuable product and still struggle if the market is too broad, the buyer is too early, the message is too diffuse, or the channel does not match how trust is built in the category. That is why the real question is not only, “Do customers want this?” The better question is whether the company can repeatedly reach the right customers, show them the right value, earn enough trust to move them forward, and expand once it is inside.
That is where Multimodal’s focus on credit unions, lending workflows, advisory boards, channel partners, and customer expansion becomes instructive. The company is not just looking for more interest; it is looking for compounding growth: more customers in the right segment, more use cases with existing customers, more partners who can surface the right opportunities, and more market trust before the sales conversation even begins.
That is the difference between traction and repeatability.
Product-market fit tells you that the product can matter; go-to-market fit helps determine whether the company can turn that value into a business.
For founders, the hard part is learning to distinguish between the signals that flatter the product and those that move the company forward. Interest, curiosity, and POCs can all be useful, but they are not the destination. The signal that matters most is the one that compounds.
NO TIME TO READ?
