By Neil Vernon, Chief Product Officer, Gresham.
For years, capital markets firms have faced the same challenge: modernising sprawling, legacy data systems. Each attempt follows a familiar pattern – ambitious platform overhauls, eight-figure budgets, years of disruption – yet the old systems often remain in use long after the new ones are live. Replacing systems rarely solves the underlying problem. Real value comes from building intelligence into existing infrastructure, incrementally, transparently and with governance at the core.
The Monolith Trap
The instinct to consolidate is understandable. If data is scattered across dozens of systems, centralising it into a single platform seems logical. In practice, these programmes often become the largest source of organisational risk. The experience of the data lake era is instructive. Many firms invested heavily in centralised repositories expecting frictionless analytics, but instead amassed uncurated, redundant data. Storage is not stewardship; putting data in one place does not make it trustworthy. Continuous curation does.Barriers to AI adoption in financial services are structural, not technological. Limited expertise, integration challenges and foundational data quality issues persist. These are problems solved by introducing intelligence thoughtfully into existing environments, not by buying bigger platforms.
Rethinking What AI can do
AI is evolving faster than many firms can track. Capabilities that were once experimental such as automated routing, data lineage inference, anomaly detection, are now increasingly reliable, persistent and able to handle imperfect data. While human oversight remains important in regulated workflows, AI can augment, and in some cases surpass, traditional manual or rules-based approaches.
Firms should approach AI with imagination and openness. Today’s models are capable of tackling tasks once considered too complex or risky for automation. This means the focus is no longer on what AI cannot do, but on how it can be applied creatively and responsibly to deliver real operational and strategic value.
APIs as the Practical Bridge
APIs are the practical bridge between legacy systems and modern intelligence. By designing an API-first architecture, firms can integrate AI services without rebuilding core platforms, allowing new capabilities to sit alongside existing infrastructure safely.
Even where APIs are absent, modern AI models can adapt, interacting directly with applications and simulating APIs to achieve the desired outcomes. Their persistence enables them to work around imperfect data, bridging gaps in legacy systems without disrupting core processes.
This incremental approach also fosters trust. By embedding controls, validation and lineage at each step, firms can continuously observe governance and ensure that data and processes remain reliable as new intelligence is deployed.
Trust is the Product, not the Platform
The ultimate output of a data strategy is trust. Business users must trust the data they make decisions on, regulators must trust reported figures and risk teams must trust that controls function as intended. Incremental approaches enable this by making governance observable at every stage.
Regulatory requirements reinforce this discipline. DORA and EMIR Refit demand operational resilience and granular validation across processes – challenges not solved by a single platform deployment but by embedded, continuous governance.
Where this Approach has Limits
Incremental deployment introduces its own complexity. Managing multiple independently deployed services requires mature DevOps, strong API governance and clear ownership. Smaller firms may find a well-implemented integrated platform more practical.
There is also a risk of fragmentation. Different business units adopting disparate tools without enterprise oversight can create silos. Incremental does not mean anarchic; clear governance frameworks and architectural principles are essential.
A Different Kind of Modernisation
The firms that will manage data most effectively in the coming years will be those that adopt a discipline of continuous, targeted improvement. This includes identifying the highest-value operational bottlenecks, deploying specific AI capabilities to address them, measuring results and iterating.
Technology delivers value incrementally in regulated environments. Start with the problem, not the platform. Deploy capabilities that prove their value quickly. Build trust through transparency and governance. Firms that succeed will not just manage data efficiently; they will make better decisions, respond to regulatory change faster and free experienced staff from repetitive manual tasks.
AI has reached a point where imagination is the only real limit. What seemed out of reach just months ago is now achievable, meaning firms can explore innovative applications that were previously considered too complex or risky. The future is open and the firms that embrace it creatively, while still safeguarding governance, will define the next era of operational excellence.
Subscribe to our newsletter
