As artificial intelligence moves from experimentation to infrastructure, investors are becoming far more selective about what qualifies as a truly AI-native company. In 2026, the gap between startups that merely integrate AI and those built entirely around it is widening fast, especially in highly regulated sectors like finance.
Dmitry Volkov, serial entrepreneur and investor, has been observing this shift from the front row. An early investor in OpenAI, Revolut, and Patreon, Volkov has deployed over $500M across more than 20 ventures and is now backing what he sees as the next logical evolution of fintech: AI-first banking.
Through his new venture, Molit.ai, Volkov is backing the development of a European bank designed from the ground up around artificial intelligence not as a feature, but as its operating system. We spoke with Volkov about what building AI startups looks like heading into 2026, how investor expectations have changed, and why he believes banking is ready for a full architectural reset.
From an investor’s perspective, what fundamentally changes when building an AI startup in 2026 compared to even three or four years ago?
One of the biggest changes is how investors look at data. A few years ago, sheer volume was often treated as a moat. From what I’ve seen, that assumption no longer holds. What matters now is whether the data is proprietary, legally exclusive, and generated through real product usage. Aggregated or scraped data is far less defensible.
Timing has also changed. Investors are no longer patient about monetisation. There is an expectation that founders understand early on how revenue will be generated. That forces teams to narrow their scope and be very precise about the problem they are solving. Broad, open-ended AI ambitions are much harder to justify today.
Competition has intensified as well. Building models is more accessible than ever, which means differentiation increasingly comes from product execution. The strongest teams I see are deeply product-driven. They focus on solving concrete user problems rather than building general-purpose models without a clear application.
You’re an early investor in companies like OpenAI, Revolut, and Patreon. What common patterns do you now recognise in startups that successfully scale in an AI-first world?
The most consistent pattern is focus. Revolut worked because it stripped banking down to what users actually needed and rebuilt the experience around that. Patreon succeeded because it addressed a very specific problem creators were facing and did so in a way that aligned incentives on both sides.
Another pattern is clarity around monetisation. The companies that scale well don’t postpone revenue discussions. They design business models that work early, which gives them flexibility later. That discipline tends to separate companies that grow steadily from those that remain stuck in experimentation.
Molit.ai is positioned as a bank rebuilt from zero with AI at its core. What convinced you that banking was ready for such a radical architectural reset?
I’m convinced this can’t be fixed by bolting AI onto legacy systems. From what I’ve seen, banks are already too constrained by how they were originally built. Their architectures were designed for a very different era, and those constraints show up everywhere.
Neobanks proved that banking is no longer about branches or paper contracts. At this point, banking is a technology and product discipline. AI has become a resource that companies simply cannot compete without. If medicine, marketing, cybersecurity, and media are all being reshaped by AI, it would be strange to assume banking is somehow exempt.
That’s exactly why we’re approaching this differently from day one. Molit.ai treats the bank itself as a technology-native system, where intelligence is embedded into the core architecture rather than layered on top.
Traditional fintech focuses on adding more features, while Molit.ai frames banking as a daily partnership with AI. How does this shift change user behaviour and long-term customer loyalty?
I think feature count is often overrated. What actually matters is how services are delivered. Most financial products force users to navigate complexity that exists for internal reasons, not user ones.
AI allows banking to happen on demand, with far less friction. A real partnership implies trust and relevance. When a system understands who a user is, what they do, and what they typically need, interactions become simpler and more timely.
Over time, that changes how people relate to financial services. Banking stops being something you manage occasionally and becomes something that fits naturally into daily workflows. That shift tends to produce stronger long-term loyalty than any single feature ever could.
Regulation and trust are major barriers in financial services. How does an AI-first banking model address compliance, security, and transparency without relying on heavy human intervention?
Being AI-first does not mean removing humans from the process. It means making human decision-making more effective. AI enables deeper investigations, stronger pattern recognition, and clearer documentation.
When designed properly, these systems are often more transparent than traditional ones. Decisions are based on broader and more consistent information, which improves auditability and accountability. In my view, this leads to stronger compliance outcomes, not weaker ones.
You’ve said that in most banks, AI acts as a barrier between the customer and real help. How did that insight shape Molit.ai’s product and interface design?
Many banks treat customer support as a cost center. Their AI systems are designed to deflect requests, not resolve them. They act more like filters than assistants.
We took the opposite approach. Assistance is built into every interaction. The system is designed to understand the customer’s history, preferences, and context so that help is relevant and timely. Instead of forcing users to adapt to the system, the system adapts to them.
Many founders still treat AI as a feature rather than a foundation. How do you evaluate whether a company is truly AI-native or just retrofitting intelligence onto legacy systems?
One clear signal is whether the system continuously learns from real usage. If customer interactions improve the product over time, that’s usually a sign of an AI-native architecture.
If AI is simply layered on top of static workflows, without influencing core logic, it’s almost always a retrofit. In truly AI-native companies, intelligence is inseparable from the product itself.
Looking ahead to 2026 and beyond, what advice would you give founders building AI-first startups today, especially those aiming to turn complex infrastructure, like banking or finance, into lifestyle products?
Founders need to be very clear about the problems they are solving and the people they are solving them for. At the same time, they need to build systems capable of adapting to problems that don’t yet exist.
AI should function as infrastructure, not as a feature. And there must be a clear path to monetisation. No matter how advanced the technology is, my experience shows me that sustainable growth still depends on understanding who pays, why they pay, and how that scales.
