At TSAM New York, MAIA will bring this discussion to life with a short demonstration of how frontier Artificial Intelligence (AI) models can already interact with MAIA, before opening the conversation to industry practitioners.
AI has advanced rapidly from experimentation towards implementation across the asset management value chain. As firms identify use cases and explore how AI could improve investment and operational workflows, attention is naturally focused on the technology itself. But there is a more fundamental question:
Is the operating environment beneath AI capable of supporting what firms are asking it to do?
For asset managers operating across fragmented platforms, proprietary systems, batch processes and multiple data sources, introducing increasingly sophisticated AI does not harmonise the underlying complexity. In some cases, it exposes it. AI may be the catalyst for change. The operating model determines how far that change can go.
AI depends on the infrastructure beneath it
The effectiveness of an AI-enabled workflow depends on its ability to access reliable information and interact with the systems where investment and operational activity take place. Where portfolio, trading, risk and operational data sit across different systems, information may need to be extracted, reconciled or transformed before it can be used. Where platforms rely on batch processes, the information available may not reflect what is happening across the portfolio now. And where integration depends on proprietary interfaces, connecting new applications can become another technology project.
This fragmentation also contributes to technology debt. Over time, maintaining legacy platforms, bespoke integrations and multiple vendor relationships can absorb an increasing share of technology budgets and internal resources. From a business perspective, that changes the role technology plays: instead of being a tangible contributor to growth, efficiency and investment outcomes, it can become a cost centre focused on maintaining the existing environment. As firms consider AI, that distinction becomes increasingly important. The objective should not be to add another layer of technology to an already complex estate, but to create an operating environment in which technology directly drives business value.
These are not new challenges. But they have become amplified when firms want AI to operate across live investment workflows. The question, therefore, is not simply whether a manager has access to sophisticated AI models. It is whether those models can access the right information, at the right time, within a sufficiently consistent and controlled environment.
From AI-ready to operationally ready
This is where the distinction between AI experimentation and operational deployment becomes important. A firm may have an AI strategy, compelling use cases and successful proofs of concept. Moving those capabilities into everyday investment operations is different.
If an AI application needs to interrogate positions, exposures, cash, P&L or trading activity, the quality of its output is inherently connected to the accessibility and reliability of that underlying information. That places greater importance on architectural principles including real-time data, interoperability, API-first connectivity and cloud-native infrastructure.
For MAIA, these are not capabilities being retrofitted in response to AI. They are embedded within the underlying technology architecture. MAIA brings mission-critical information, portfolio management, trading, risk, compliance, IBOR and middle-office workflows into a unified environment, removing fragmentation and creating consistency across the investment lifecycle.
That architecture also creates a controlled foundation for AI to interact with investment data today. Through MAIA’s permission-aware APIs, users can leverage frontier models such as Claude and Codex to interact with real-time investment information within the permissions of the authenticated user. This provides a governed route for AI to support activities such as querying data, explaining information and drafting reports, without giving the AI layer unrestricted access to the underlying investment environment.
Importantly, this does not require AI to become the source of truth. MAIA’s IBOR remains the underlying data aggregation and deterministic calculation layer, with positions, cash, exposures and P&L calculated within the platform. AI can sit above that foundation as a query, explanation and report-drafting layer. The distinction matters: rather than asking AI to replicate critical business calculations, firms can use it to interact with trusted information produced by the investment infrastructure beneath it.
We are already seeing several use cases where the MAIA API provides the gateway for AI to interrogate exposures, explain P&L movements, summarise cash and margin pressures, triage operational exceptions or draft investor reporting commentary. In each case, the business benefit is not that AI replaces the underlying investment infrastructure. It is that users can access trusted information faster, understand it more clearly and act with greater confidence.
From ambition to operational reality
Over time, AI is expected to become an operational command layer for asset managers, helping teams prioritise exceptions, explain breaks, anticipate cash or margin pressures, monitor close processes, draft reporting commentary and surface emerging risks. Scaling the use of AI beyond individual use cases will hinge on strong control frameworks, clear governance and permissions, trusted data and well-defined workflow ownership.
