Spend enough time talking to people who run innovation inside large financial institutions and you start to notice the same pattern. It’s never spelled out in strategy decks. It rarely appears in public statements. But it shapes everything these organisations can and cannot do with AI.
There are only two real strategies.
Either a company believes it should grow all of its AI capabilities in-house.
Or it believes it should partner with the world.
Everything else is just a variation of these two instincts.
The internal-first strategy is the most intuitive. These institutions operate under heavy regulation. They have auditors, risk committees, and governance structures designed to minimise surprises. So they do what feels safest: train internal teams, keep control of the stack, and avoid dependency on outsiders. The people inside these organisations are incredibly smart. But the system learns at the speed of its own hierarchy. Even highly talented teams struggle to break out of predefined roadmaps or bring in unconventional ideas. And because roles must map neatly onto internal structures, frontier talent often doesn’t have an obvious place to land.
The ecosystem strategy looks almost opposite. These institutions accept that no internal team—no matter how strong—can keep up with a rapidly evolving frontier. So instead of playing defense, they open the windows. They run co-creation trials. They experiment with small, messy POCs. They invite in specialists who don’t fit into their job families but who expand their field of vision. These organisations tolerate ambiguity because they understand that early-stage innovation is inherently ambiguous. They value learning over control.
Neither strategy is inherently better. But they produce very different cultures, speeds, and outcomes.
This distinction matters enormously for founders like me. When you build something interdisciplinary—something that blends agentic AI, causal inference, actuarial logic, sustainability, and financial modeling—internal-first organisations may love the idea but simply cannot absorb you. Not because they don’t want to, but because their structure doesn’t have a slot for the kind of intelligence you bring. Conversations that feel promising quietly stall. Projects die in committee. The door closes without ever looking like it was open.
Ecosystem-style institutions behave differently. They don’t ask “Where do we put you?” They ask “What can we build together?” They don’t expect certainty at the beginning. They expect exploration. And they treat startups not as vendors but as thought partners.
Once you see this distinction, it becomes liberating. You stop trying to push into institutions whose evolutionary path makes them rigid by design. You start focusing on the ones whose strategy aligns with your own—institutions that evolve through curiosity, not through control.
In the end, AI doesn’t transform organisations. It reveals them. And the smartest thing a founder can do is work with the ones whose identity matches the future they want to build.
