AI in Investment Operations: Process Intelligence First, Then Investment Intelligence

Introduction

When discussing AI in investment management, the focus is often on the investment decision itself. For investment operations, however, a different question is more pertinent: where can AI be deployed in such a way that accountability, governance and traceability are maintained?

The answer often lies both before and after the Investment Committee. Incoming documents must be recorded and categorised, process cases prepared, approvals managed and decisions documented. It is precisely here that it is determined whether an AI-driven initiative will translate into a robust operational contribution.

Process Intelligence rather than decision automation

Investment Intelligence aims to provide insights for investment decisions. Whilst this may be a relevant area, it is not the only – and in many organisations not the primary – starting point for AI.

Process Intelligence, in this context, describes the ability to support operational workflows based on their context: assigning information to a process case, preparing verification steps, highlighting outstanding issues and managing documentation in line with clear roles and approvals. The technical assessment and the decision remain with the responsible individuals and committees.

This distinction is important. AI can structure content or provide guidance. However, it should not obscure who made, reviewed or approved a decision. For regulated and institutional environments, this transparency is not an add-on, but part of a robust process design.

Four operational starting points

1. Intake as a structured entry point

In investment operations, information often arrives via various channels: documents, emails, spreadsheets or reports. Before this can be turned into a processable task, it needs to be categorised and assigned. AI can support this initial stage by pre-sorting content and preparing it for further review. The definitive assignment to a task and the professional assessment remain controlled work steps.

2. Review with clear accountability

Reviews improve in quality when requirements, responsibilities and supporting evidence are not managed separately. AI can flag missing information, discrepancies or outstanding queries. The benefit arises not merely from the flag itself, but from the subsequent process: Who carries out the review? According to which criteria? What decision was made? Where is the supporting evidence?

3. Approvals as traceable process steps

Approvals are more than just a change of status. They link roles, mandates, decision-making powers and a specific process case. An AI-supported workflow can prepare information for approval or check for completeness. The approval itself, however, requires clearly defined responsibilities and a documented decision.

4. Documentation generated during execution

Traceability becomes robust when it is not simply added at the end of a process. It should be generated during processing: from roles, tasks, approvals and the associated information. In this way, documentation becomes the result of execution rather than a separate post-processing step.

What this means for process architecture

AI does not deliver its benefits in investment operations in isolation. It requires a process framework that links business objects, roles, responsibilities and evidence. Without this framework, an AI result remains difficult to attribute: it is unclear to which case it applies, who reviews it and how it feeds into a decision.

Such a process architecture links processes, roles, responsibilities and documentation to the relevant business objects. This applies, for example, to governance and compliance processes, asset management and fund administration, as well as approval and decision-making processes. For AI-supported operations, such a structure can provide a controllable framework within which recommendations are processed, decisions are assigned and evidence is recorded directly during execution.

This is not a statement about an automated AI decision-maker. It is a design principle: first clarify the process, responsibility and evidence; then deploy AI in a targeted manner in defined steps.

Conclusion

The first robust application of AI in investment operations need not begin with an automated investment decision. It can start with the recurring operational tasks that simultaneously shape quality, speed and governance: gathering, verifying, approving and documenting information.

For institutional real-assets organisations, the challenge therefore lies in the interplay between AI and process architecture. AI can prepare and structure work. Responsibility for decisions, traceable approvals and documented processes remain the foundation.