Broadridge’s recent announcement that its agentic AI is live in production across more than 40 clients signals a move from capability disclosure to commercial commitment. The firm is now offering two consumption models: a managed services arrangement where Broadridge runs operations end-to-end with the agentic layer on top, and a standalone deployment of the same platform into a firm’s own infrastructure via open-standard APIs. The headline figure attached to the managed services path is a 30% Day 1 reduction in operational costs, with further shared savings as the AI matures.
The productisation builds on Broadridge’s investment in agentic AI specialist DeepSee in January and a sequence of capability disclosures since. Underneath the agents, though, sits a more consequential argument about data, with Broadridge positioning the data layer as the foundation that separates production-grade agentic deployment from experimentation.
The Ontology as the Load-Bearing Claim
Broadridge describes its financial services ontology as “completed” – a notably strong claim for a data model in financial services. Asked what completeness actually means in practice, Tom Carey, President of Global Technology & Operations at Broadridge, points to the firm’s regulatory reporting data as the original starting point – the highest-volume, fastest-flowing data Broadridge consumed, and one already normalised across multiple jurisdictions.“We stepped back around five years ago, in 2020, and said to ourselves: given the amount of data we hold on behalf of our clients, and given the breadth of our services, we’re well placed to define this for the market going forward,” he tells TradingTech Insight. “We’ve been able to map all of Broadridge, which is a very wide estate, and the ontology has only increased by about 25% in size since day one. That gives you a good picture: from a mapping perspective across equities, fixed income, repo, derivatives, the model is extendable and flexible.”
The 25% figure is a substantive claim. A schema that grows by only a quarter while being extended to cover the firm’s full asset class and jurisdictional reach suggests the model is mapping new instruments into existing structures rather than appending new ones, and that matters when an agent has to reason consistently across asset classes.
The argument Carey makes is structural. Agents built on inconsistent data, or on point solutions stitched across fragmented systems, become hard to maintain, harder to audit, and progressively less trustworthy as scope grows. A single normalised data model removes that drag. Whether Broadridge’s ontology proves to be the right one, or whether competing models from FINOS, ISDA CDM and the open finance data community gain ground over time, is a question the market will work out. What Broadridge has made clear is that it sees the data layer, not the agent layer, as the moat.Two Consumption Paths, One Commitment
The two-path commercial model is straightforward in shape. Firms that want full operational transformation can hand the workflows over to Broadridge as a managed service, with Broadridge’s people, technology and agentic layer combined under a single contract. Firms that want to keep operations in-house can deploy the platform into their own infrastructure and run it themselves, with the same underlying ontology and the same agent capabilities accessible through open-standard APIs.The 30% Day 1 cost reduction figure is central to how the offering is positioned commercially.
“We measure that against a TCO based on the client’s current operational spend for the in-scope activities, plus their IT spend for the same in-scope activities,” says Carey. “Historically, a lot of firms have gone with an incremental savings model – 5% this year, 15% the year after, and so on. Our belief is that the world has evolved and reached a point where firms recognise the AI is coming faster, so we’re prepared to commit to those savings up front for a term-based agreement.”
For firms taking the standalone deployment path, the savings profile is necessarily different. Broadridge is supplying the platform but not the operational scale, and Carey makes no equivalent up-front commitment on the in-house route, a meaningful distinction for any firm currently weighing build-versus-buy on agentic operations.
What Is Actually in Production
The capabilities Broadridge lists as live are the ones that have featured in its disclosed roadmap over the last 18 months: automated trade fails management and break resolution, account opening and maintenance, real-time valuation exception handling, customer inquiry automation, and email workflow processing in partnership with DeepSee. The more interesting question is how autonomous any of those agents actually are, and Carey is more measured than the broader industry framing around fully autonomous AI.
“If you take an entire process – the whole thing running with no human involved – we don’t believe in that, and we wouldn’t advise it,” he says. “Clients would question what we were doing without our risk and controls. But within the flow, anything you do in operations is typically a multi-step process, and agents are doing entire steps in that process. Anything that breaks our rules or thresholds, or requires further validation, the human comes back in. We don’t believe in the machine running the entire estate – that’s probably a step too far.”
The operating model that emerges is a graduated one. Agents handle entire steps within multi-step workflows. Genuine straight-through transactions – those validated against the firm’s rules and risk thresholds – flow without human intervention. Anything that breaches a threshold, or carries enough value or risk to warrant human review, surfaces to a supervisor with the relevant context already assembled on a single screen. That is probably closer to what the institutional buyer side wants in practice than full autonomy would be in any case.
Curating, Not Open-Sourcing
Broadridge has flagged that it is exploring making core elements of the ontology available as an open industry resource. Carey’s position, when pressed, is more bilateral than that phrase might imply. Broadridge intends to keep curating the ontology itself, at least for now, and is not in active discussions with FINOS or any other neutral body about transferring stewardship.
“At the moment, we would continue to curate it. We’re not against doing something different, but our initial view is that we would curate it ourselves,” says Carey. “The bigger use case is making the ontology, or a domain of the ontology, available for clients to consume, allowing them to build other applications against it. It’s not our data – it’s our clients’ data. We augment it, we add information to it, but we believe that data is owned by the client.”
In practice, that points to a published reference data model for firms that lack the scale to build their own, made available so that data can flow back to Broadridge in a consistent format. It is a more measured proposition than full open-sourcing would imply. Whether the position evolves over time, and whether Broadridge moves towards a more openly governed model as the agentic platform gains adoption, will be worth watching.
The Implication for Build-Versus-Buy
Beyond what is already live, Carey sketches a near-term pipeline of further agents and a longer-range ambition that pushes Broadridge’s agentic thinking beyond operations and into the workflows themselves.
“We’ll come to market with incremental rollouts of new agents over the next six to nine months. We have a whole set in our test environments going through final testing,” says Carey. “We have bigger concepts around design as well – how we can actually automate the design of future workflows, allowing clients and ourselves to design workflows and deploy them safely. We’re evolving to the next level, allowing our senior leadership and our operations leadership to design going forward.”
For buy-side and sell-side firms, the question this raises sits one level deeper than the agent layer. If the ontology argument holds – and the 25% expansion claim is the strongest evidence offered that it does – the build-versus-buy decision begins to extend to the data model the agents run on, not just the agents themselves. That is a different kind of commitment, and one that puts Broadridge in a different competitive frame from the agentic point-solution providers it currently sits alongside.
Subscribe to our newsletter
