Banks are splitting into two AI classes
Why your infrastructure decides your AI future
Insights by Zaf Argyropoulos
21 July 2026
Executive summary
Banking is experiencing a two-speed AI divergence, with large institutions deploying revenue-focused initiatives while smaller banks are still experimenting with basic use cases like summarising documents and researching. This structural divide exposes smaller institutions to critical operational risks, demanding to move away from bolted-on AI layers. Mambu's Intelligent Core enables agentic operations at scale, giving AI agents the real-time data access and decisioning power they need to operate autonomously within the core itself.
_____________________
Banking's two-speed AI race
Everyone is talking about how quickly AI is advancing. But after sitting down with our customers, industry experts, and AI founders, I wanted to look deeper at a more pressing question: where is the value actually concentrating, and who is going to capture it?
What I notice is that financial institutions are diverging sharply on how they adopt AI. Tier 1-2 institutions are moving with conviction. They have already bypassed pilots into executing against specific value targets. These institutions also expect most of that value to come from revenue-focused initiatives rather than cost-cutting measures.
Meanwhile, tier 3 and smaller banks are largely still experimenting with AI tools. Their AI activity is focused on more basic use cases such as, summarising documents and doing research. While trying to address internal risk and compliance concerns, they aim to deploy agentic solutions in the next 18-24 months.
A structural divide in vendor strategy
The second point that came up repeatedly in my conversations was that the larger banks are building their own model-agnostic AI infrastructure and agentic workflows. They are delivering value directly across the banking stack, starting from the back and middle office.
Smaller institutions don't have that option. They're working with tighter budgets and smaller teams and don’t have capacity and resources to own capabilities that are not critical to them. Instead, they predominantly look to buy solutions that are available in the market and help them address internal deficiencies.
The sovereign AI inflection
One thing I didn't expect to come up as much as it did - is sovereign AI. While larger banks can invest in locally-hosted alternatives, smaller institutions don’t have that potential. Their dependency on non-European LLM providers exposes them to operational risk. What happens to a retail bank that shifts its call center from humans to AI agents and there is a sudden price hike? How can a commercial bank compensate for top-tier model restrictions if those are vital in rewriting their codebase or enabling its banking operations?
The infrastructure gap is the strategy gap
When you tie all of these threads together, everything comes back to architecture. While workarounds can work in the short term, the institutions moving fastest are the ones whose core systems can actually support rapid change. Banking at the speed of AI requires a true SaaS foundation. It needs clean accessible data, APIs with streaming notifications, accessible via MCP server so agents can act without rigid integrations, and orchestration capability.
By putting all that together the core transforms from a system of record to a system of engagement, enabling human operators to complete repetitive tasks and get answers within seconds to questions like "Help me summarise a client's position with the bank as I'm about to speak with them about their banking needs" or "Why was customer X charged the fee last week?"
For smaller banks to survive the AI race, they must bridge the infrastructure gap with guided composability, choosing a foundation built for continuous evolution. Banking at the speed of AI starts with building on a foundation designed for it.
This is exactly why Mambu developed Intelligent Core. Instead of forcing you to build a messy, separate AI layer bolted on top of legacy tech, Mambu provides a native banking infrastructure that can receive, interpret, and act on AI signals in real time.
Zaf Argyropoulos, Senior M&A Director
Zaf works closely with our commercial and product teams, as well as being in regular touch with customers, other financial institutions and industry experts to better understand industry trends, customer needs and devise strategies to address those.
Curious how Mambu’s Intelligent Core can support your business goals?
Get in touch with one of our experts for a conversation
© Mambu 2026