The findings from our recent Datos Insights report, Agentic AI In Capital Markets, are clear. The mandate for agentic AI is real. Budgets are moving. Board-level commitment is present.
What’s less clear is whether firms have resolved the legacy integration and data-quality challenges required for scale. Even less clear is whether they recognize that the governance foundation is already in place to support it. The evidence on the first question is mixed. On the second, most firms have not yet connected the dots.
The real advantage for capital markets firms lies not in what they’re missing but instead in what they’ve already built. A decade of regulatory pressure has constructed the exact governance scaffold that production-scale agentic AI requires.
The Governance Inheritance Advantage
A decade of regulatory pressure (SR 11-7, MiFID II, FINRA Rule 3110) forced capital markets firms to build the exact governance infrastructure that agentic AI requires: model validation frameworks, audit-trail architecture, and human-review checkpoints. These were compliance mandates, not strategic choices.
Most capital markets firms face renovation, not construction: The framework and discipline exist, and adapting them to handle LLMs and multi-agent architectures is a different order of effort than building governance from zero on unmarked regulatory ground. The advantage isn’t automatic; it depends on treating existing infrastructure as a foundation rather than overhead.
Where the Real Mandate Shows Up
Two-thirds of firms expect AI spending to grow more than 10% over the next two years. But the distribution across the operational stack is what matters. Capital markets firms are investing simultaneously in investment research, surveillance, reconciliation, document management, and workflow automation. Cloud migration and DLT never achieved this breadth. This concurrent investment across front, middle, and back office isn’t what a line-item technology initiative looks like.
More than a third of firms are committing $10 million or more to surveillance, investment research, and workflow automation. That spending level signals a shift from GenAI co-pilot experiments to agentic-level commitments: multistep workflows coordinating across systems with human review at defined checkpoints, not every transaction.
The Barriers Are Solvable; the Sequence Isn’t
Capital markets AI adoption faces five primary barriers: legacy infrastructure, model governance, talent availability, ROI measurement difficulty, and data quality. None is new. All have been solved elsewhere. The three most common deployment failures—legacy integration surprises, governance blockers during validation, and late-stage data-quality discoveries—are preventable. They’re expensive to fix mid-program and free to address as prerequisites.
Most firms already have the foundation for the two biggest barriers: legacy integration and model governance. Bloomberg, BlackRock’s Aladdin, and SimCorp connect broadly across capital markets systems. Nearly all firms operate at least one. The faster path is treating these as the organizing layer for AI deployment, not building new infrastructure alongside them. Capital markets firms have also spent a decade building validation discipline under regulatory compulsion. The advantage is using that framework as your AI governance foundation, not rebuilding from scratch. Formalizing this extension before production deployment, not after, separates firms that will deploy successfully from those that become blocked during validation.
Firms that leverage what they already have—starting with compliance automation where governance exists—will move fastest. Shortcutting that foundation slows everything else.
Next Steps
Capital markets firms have a structural advantage in the transition to agentic AI that no other financial services segment possesses. The Datos report details a three-stage execution pathway: Start with risk surveillance and compliance (where governance already exists), build ROI metrics and cross-system data integration through operations and reconciliation, then move to investment research and portfolio workflow integration—and the sequence isn’t optional, as each stage earns the infrastructure for the next.
Big-four consulting firms dominate AI advisory relationships because most firms are still in the early stages of planning. That’s appropriate for now, but the consulting mix will need to shift as programs move to implementation and production. Banks and other capital markets firms should build that transition plan into AI program design immediately. Strategy firms are right partners for initial architecture and readiness assessment. However, they are not the right partners for scaled implementation and production deployment. Firms that treat those relationships as permanent rather than phased will find themselves without the specialist talent they need when it matters most.
But all of that execution—the sequencing, the vendor transitions, the infrastructure decisions—rests on one thing: Governance is foundation. It’s that simple. Everything else follows from recognizing it.
Datos Insights is the leading research and advisory firm serving the financial services industry. To learn more, visit datos-insights.com.
