Over the past two years, most AI conversations have focused on models — how powerful they are and how quickly they’re evolving. What tends to get less attention is the condition of the data those models depend on.
In the financial services industry, data has always been central to how we manage risk and generate insights. That hasn’t changed. What AI has changed is how quickly gaps in a data strategy become visible. If ownership is unclear or definitions vary across systems, AI tends to surface those issues rather than resolve them.
That’s why, for BlackRock, AI strategy starts with something more fundamental: data ownership and meaning.
In many organizations, data accumulates as a byproduct of applications and workflows. This exhaust is stored, transformed and reported on, but over time context erodes, definitions drift, and accountability becomes blurred. That can work when experienced analysts are interpreting results and filling in gaps from institutional knowledge, but it doesn’t translate as well when machines are expected to reason over the same assets.
If AI systems are going to operate reliably, the data they consume needs clear ownership, documented lineage and consistent definitions. We’ve increasingly treated data as a product rather than as exhaust. That means someone is accountable for it, governance is built in from the beginning and data is designed so it can be reused — not just for a single report, but across all workflows, including AI-driven ones.
One decision that has shaped this work is our choice to standardize our data estate without centralizing.
In regulated industries, the traditional response to complexity has been consolidation — move everything into one warehouse and control it there. That can improve oversight, but it can also slow teams down and create distance between data and the people who understand it best. We operate in a federated model where business and product teams remain responsible for their data products because they’re closest to the investment processes those data sets support.
Snowflake has been foundational to how we’ve standardized those capabilities by giving us a consistent way to secure, share and audit data across the firm. Strong role-based access controls and native data sharing allow us to scale while staying interoperable without requiring every decision to flow through a single centralized group.
You can see that balance in how we deliver data through the Aladdin Data Cloud. Not long ago, distributing data meant flat files, FTP transfers and large schema definitions. It worked, but it wasn’t built for today’s environment. Now clients can integrate their books of record and market data directly into their broader data estates in a modern, cloud-native way.
In financial services, governance can’t be layered on after the fact. As we’ve continued to evolve our data platform, we placed significant emphasis on metadata and accountability. No asset comes onto the platform without clear ownership and documentation of its provenance. As AI-driven workflows and natural language interfaces become more embedded in day-to-day processes, that traceability only becomes more important.
AI also surfaces something more subtle: semantics. Two systems can use the same field name and mean completely different things. Humans can often work through that, but AI systems can’t. If context is implicit rather than explicit, reliability breaks down quickly.
That’s one reason I’m particularly engaged in the industry’s work around the Open Semantic Interchange (OSI). The objective is straightforward — enable organizations to exchange not just data, but the meaning behind it. When semantics are portable, you don’t have to rebuild interpretive layers every time you introduce a new tool, model or agent. In an AI-enabled ecosystem, that kind of interoperability becomes practical very quickly.
As we’ve strengthened ownership and shared standards, we’ve seen a noticeable shift in how teams operate. There’s less time spent managing bespoke transfers or stitching systems together and more time focused on improving the clarity and usefulness of data products themselves. In our experience, agility and control aren’t at odds. With clear standards and accountability, they tend to reinforce each other.
Markets and technologies will continue to evolve. New asset classes will emerge. Interfaces to data will become more intuitive. Preparing for that future is about continuing to mature the fundamentals — ownership, governance and clear meaning in the data itself versus chasing the next model release.
Our mission at BlackRock is to help more and more people experience financial well-being. AI can accelerate how we deliver on that mission, but only if the data underneath it is owned, governed and clearly understood.
For organizations thinking about AI strategy, that’s where the real work begins.
