Can AI really understand corporate actions?

Few functions within asset servicing remain as dependent on human interpretation as corporate actions. Every dividend, merger, rights issue, or tender offer begins with an issuer announcement that must be interpreted before it can translate into operational instructions. Unlike settlement or reconciliation, the challenge is not simply processing transactions faster – it is understanding what those announcements mean. Issuer documents arrive in multiple formats across different jurisdictions, and data vendors can interpret the same event differently, meaning operations teams must determine which interpretation is correct before any instruction reaches a client.

Large language models (LLMs) and generative AI are beginning to change that. Unlike traditional rules-based automation, AI can analyse unstructured documents, extract key information, and summarise complex events in natural language – bringing a new layer of interpretation capability to one of financial services’ most demanding operational functions. Corporate actions, however, present one of AI’s toughest tests: a misinterpreted election option or payment date can have immediate financial consequences, meaning firms must balance efficiency with operational control.

Smartstream has moved beyond proof of concept, applying AI to review narrative documentation, summarise event descriptions, and identify conditional language within lengthy issuer announcements. Adam Cottingham, Product Manager for Corporate Actions at Smartstream, explains the core value: “Complex corporate actions often come with detailed offer documentation, which is not written to a standard and is often perceived differently by an individual. AI provides the ability to move unstructured data to a structured transformation, providing evidence on the lineage at scale, extracting the pertinent information, and removing the noise.”

Rather than replacing existing deterministic processing, AI acts as an interpretation layer above it. Rules continue to govern entitlement calculations and validations, while AI focuses on understanding documentation that previously required manual review. Cottingham describes this as a shift in classification logic: “AI gives the ability to move from deterministic classification to learned, probabilistic classification, focusing the user on the most important tasks.” Smartstream’s Smart Actions end-to-end corporate actions automation platform brings this capability into a fully governed, auditable workflow environment covering all mandatory, mandatory with choice, and voluntary ISO event types.

For complex events such as bankruptcies, tender offers, mergers, and restructurings – involving hundreds of pages of legal documentation and evolving terms – AI’s greatest value lies in accelerating understanding rather than replacing judgement. Cottingham is direct about where that boundary falls: “AI can realistically support these events, but the value clusters on the information side, and the trust boundary runs between helping an operational user think and being allowed to decide.” Entitlement calculations, proration methodologies, and election decisions remain deterministic processes that continue to require rules engines and human oversight. As Cottingham puts it: “The single biggest benefit is AI as a reading accelerator.”

Transparency determines whether AI delivers genuine efficiency or simply relocates manual work. “Being able to see why the AI produced a value and trace it to a source is what lets a reviewer validate an output in seconds instead of re-deriving it,” Cottingham explains. “Without this, AI merely relocates the manual re-checking rather than removing it.” Operations teams are evolving accordingly – moving, in Cottingham’s words, “from the doing to supervising”, with expertise shifting towards ambiguous events, complex terms, and fiduciary decisions where accountability, judgement, and oversight remain inherently human responsibilities.

AI’s effectiveness in corporate actions also depends on the quality of the underlying data – a longstanding industry challenge. Cottingham identifies a specific risk that firms must manage: “AI’s risk in this process is not dirty or incomplete data. It is making incorrect data look clean.” This reinforces the importance of robust operational controls and industry standardisation initiatives such as ISO 20022, which reduce the volume of interpretation required before information reaches downstream systems. On the regulatory front, Cottingham sees the EU AI Act as a governance enabler rather than an obstacle: “Most asset-servicing AI sits outside the high-risk regime. The EU AI Act is likely to prove less restrictive than feared and more useful than expected.” He argues the legislation helps move organisations from questioning whether AI can be trusted to establishing governance frameworks that allow firms to formally sign off on AI outputs. For asset servicers pursuing production AI, building that confidence – across data quality, explainability, and operational controls – is becoming as much a governance challenge as a technological one. To find out how Smartstream is developing AI capabilities across the transaction lifecycle, connect with the Smartstream team.