.jpg)
Can AI really understand corporate actions?
08 Jul 2026
AI is beginning to reshape corporate actions processing, but as firms move from experimentation to production. Can AI be trusted to interpret legally binding issuer announcements while complying with emerging regulation?
Image: raimundas/stock.adobe.com
Reading between the lines
Few functions within asset servicing remain as dependent on human interpretation as corporate actions. Every dividend, merger, rights issue, tender offer, or bankruptcy begins with an issuer announcement, legal notice, or market communication that must be interpreted before it can be translated into operational instructions. Unlike settlement or reconciliation, corporate actions have resisted full automation because the challenge is not simply processing transactions faster but understanding what they mean.
Issuer announcements arrive in multiple formats across different jurisdictions, while data vendors can interpret the same event differently. Before an instruction reaches a client, operations teams must determine which interpretation is correct.
AI, particularly large language models (LLMs), is beginning to change that. Unlike traditional automation — which relies on predefined rules — generative AI can analyse unstructured documents, extract key information, and summarise complex events in natural language. According to Eurofi, 91 per cent of financial services firms are either assessing AI or already using it in production, with document analysis and knowledge management among the fastest-growing use cases.
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.
Dirk Löscher, head of custody and investor solutions at Clearstream, believes corporate actions are both an ideal use case and one of AI’s greatest challenges.
“Corporate actions are a natural fit for AI because the process still depends heavily on interpreting unstructured issuer announcements, agent emails, and legal documentation,” he says. “At the same time, that is exactly what makes the area challenging: the data is often complex, market-specific, and not always presented in a consistent format.”
While AI can extract and structure information more quickly than traditional workflows, he stresses that “it must operate within strong controls because the economic impact of a wrong interpretation can be material.”
That balance between efficiency and control has become the defining theme of AI adoption across corporate actions.
From proof of concept to production
For several years, AI discussions within financial services centred largely on experimentation. Today, firms are beginning to deploy AI in production, focusing on the most manual parts of the corporate actions lifecycle rather than attempting to automate the entire process.
One of the clearest applications is document interpretation.
Corporate actions teams receive issuer announcements in PDFs, emails, HTML files, and lengthy legal documents that must be converted into structured operational data. Traditionally, this required analysts to manually identify key dates, election options, and event conditions before processing could begin.
According to Adam Cottingham, product manager for corporate actions at SmartStream, AI is proving particularly effective at bridging the gap between unstructured documentation and structured workflows.
“Complex corporate actions often come with detailed offer documentation, which is not written to a standard and is often perceived differently by an individual,” he explains.
“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 automation, AI is becoming an interpretation layer that sits above deterministic processing.
Rules continue to govern entitlement calculations and validations, while AI focuses on understanding documents that previously required manual review.
Löscher describes this as moving workflows “from manual reading, interpretation and re-keying towards assisted extraction, classification, and preparation of structured event data”.
“The objective is not to remove controls, but to reduce manual effort, improve consistency, and increase operational resilience before an event reaches the validation stage,” he informs.
Both firms are now seeing these capabilities move beyond proof of concept.
Clearstream has launched projects to automate asset servicing events, beginning with voluntary repurchase offers before expanding into additional event types.
“This is not AI for its own sake,” notes Löscher. “It is targeted automation of a real operational pain point.”
SmartStream, meanwhile, is using AI to review narrative documentation, summarise event descriptions, and identify conditional language hidden within lengthy issuer announcements. These capabilities remain paired with human validation, allowing operations teams to focus less on locating information and more on verifying it.
AI is also beginning to reshape workflow management. Rather than processing every event in arrival order, firms can prioritise work based on confidence levels and operational risk.
“AI gives the ability to move from deterministic classification to learned, probabilistic classification,” comments Cottingham, “focusing the user on the most important tasks.”
When complexity wins
While AI performs well on routine events, its limitations become clearer as corporate actions grow more complex. Bankruptcies, tender offers, mergers, and restructurings often involve hundreds of pages of legal documentation, multiple election options and evolving terms. Operations teams must not only extract information but interpret legal language, understand market practice, and ensure operational instructions accurately reflect an issuer’s intentions.
Here, AI acts as an assistant rather than a decision-maker.
“Rules-based automation performs well when the input is stable and structured,” observes Löscher. “Corporate actions rarely offer that luxury.”
Instead, he believes AI works best as an interpretation layer, combining document analysis with deterministic validation, and operational review.
Cottingham agrees, arguing that AI’s greatest value lies in accelerating understanding rather than replacing judgement.
“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.”
AI can assemble information from lengthy merger agreements, summarise restructuring documents, compare amended announcements, and identify changes to election deadlines far more quickly than manual review.
“The single biggest benefit,” he adds, “is AI as a reading accelerator.”
However, entitlement calculations, proration methodologies, and election decisions remain deterministic processes that continue to require rules engines and human oversight.
As Löscher concludes: “The right model is not blind automation; it is controlled augmentation.”
The human in the loop
While AI is becoming increasingly capable of interpreting corporate actions, technology providers agree that it should remain accountable to human expertise.
Rather than replacing operations teams, AI is reshaping their role. Historically, specialists spent much of their time reading lengthy documents, comparing vendor feeds, and validating routine events. As AI absorbs these repetitive tasks, operations professionals are increasingly focusing on supervision, exception management, and quality assurance.
Löscher expects that trend to continue.
“In the near term, AI should primarily act as an assistant to operations teams: extracting information, preparing event data, reducing repetitive work, and surfacing exceptions,” he predicts. “Over time, more deterministic parts of the workflow may become highly automated, but decision making in complex or judgement-based scenarios should remain under human oversight.”
Cottingham shares that view, arguing that operations teams are moving “from the doing to supervising”, with expertise shifting towards ambiguous events, complex terms, and fiduciary decisions. While AI can accelerate document analysis, accountability, judgement, and oversight remain inherently human responsibilities.
That human oversight is closely linked to one of the industry’s biggest concerns — explainability.
Unlike traditional rules-based automation, LLMs can reach conclusions without clearly showing how they arrived there. For corporate actions — where every decision must be auditable — that creates an obvious challenge.
“Corporate actions processing takes place in a control-heavy environment, and users must be able to understand why a data point was extracted, where it came from and how it was transformed,” informs Löscher. For Clearstream, trusted AI requires visible lifecycle tracking, validation, regression testing, and clear ownership.
Cottingham agrees that transparency determines whether AI delivers genuine efficiency.
“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,” he says. “Without this, AI merely relocates the manual re-checking rather than removing it.”
As a result, firms are increasingly combining AI with confidence scoring, audit trails, and human approval workflows. Rather than replacing operations professionals, AI is becoming another participant in the workflow — one whose outputs are reviewed, measured, and continuously refined.
The future role of operations is therefore shifting from processing information to governing how intelligent systems interpret it.
Trust by design
As AI moves into production, the conversation is shifting from capability to governance. Financial institutions are increasingly focused not only on what AI can do, but on whether it can be deployed in a way that is transparent, explainable, and accountable.
That discussion has gained momentum with the introduction of the European Union’s AI Act. Although most AI applications used in corporate actions are unlikely to fall within the legislation’s high-risk category, the Act reinforces principles such as human oversight, transparency, and documentation that firms are already building into their AI strategies.
Cottingham believes the regulation will encourage confidence rather than hinder innovation.
“Most asset-servicing AI sits outside the high-risk regime,” he notes. “The EU AI Act is likely to prove less restrictive than feared and more useful than expected.”
He argues that the legislation helps move organisations from questioning whether AI can be trusted to establishing governance that allows firms to “sign off on it”.
Löscher also sees regulation as an enabler rather than an obstacle.
“Regulation will raise the bar, but that is not necessarily a negative,” he states. “In asset servicing, trust is built through governance, controls, transparency, and accountability.”
He expects the AI Act to encourage more disciplined model design, testing, monitoring, and documentation. “That may slow down superficial adoption, but it should accelerate sustainable adoption where AI is deployed with the right safeguards.”
Ultimately, firms recognise that confidence in AI cannot rest solely on accurate outputs. Models must be consistently explainable, auditable, and supported by clear accountability. In that sense, successful AI adoption is becoming as much an exercise in governance as it is in technology.
Better AI starts with better data
While generative AI is transforming how firms process unstructured information, its effectiveness still depends on the quality of the underlying data.
Corporate actions have long been plagued by fragmented information, with issuer announcements arriving in different formats and multiple vendors often interpreting the same event differently. AI can help identify inconsistencies more quickly, but it cannot resolve poor-quality data on its own.
“AI can help identify inconsistencies, missing fields, and unusual data patterns earlier in the lifecycle,” says Löscher. “But it does not magically solve data quality. The value comes from making uncertainty visible: flagging gaps, preparing structured outputs, and routing exceptions to the right operational experts for validation.”
Cottingham agrees, arguing that AI is particularly effective at detecting discrepancies by comparing multiple sources simultaneously. However, he cautions against assuming AI-generated outputs are automatically more accurate.
“AI’s risk in this process is not dirty or incomplete data,” he explains. “It is making incorrect data look clean.”
For both firms, AI reinforces the importance of industry standardisation rather than replacing it. Initiatives such as ISO 20022 and broader efforts to create structured, machine-readable corporate actions data will reduce the amount of interpretation required before information reaches downstream systems.
Löscher believes AI’s greatest contribution may be exposing weaknesses that have always existed. “If the market succeeds in establishing the CSD as the central starting point of a high-quality data flow, AI can help convert that information into structured, machine-readable data at scale.”
Ultimately, AI is proving to be as much a mirror as a solution. Rather than fixing fragmented data, it is highlighting where better standardisation and digitalisation are still needed — making data quality, rather than AI capability, the industry’s biggest determinant of success.
Beyond automation
For decades, the corporate actions industry has pursued greater automation.
Straight-through processing (STP) reduced manual intervention. Standard messaging formats improved interoperability. Workflow technologies increased operational efficiency.
AI represents the next stage of that evolution — but not because it replaces people.
Its significance lies in addressing something previous technologies could not: interpreting information before it becomes structured data.
The technology is already demonstrating measurable value.
It is reducing the time required to analyse complex documentation, highlighting discrepancies between information sources, prioritising operational exceptions, and allowing specialists to concentrate on the events that genuinely require experience and judgement.
At the same time, the industry’s early deployments reveal equally clear boundaries.
Corporate actions remain rooted in legal interpretation, operational accountability, and fiduciary responsibility.
Those are areas where transparency, explainability, and human oversight are likely to remain essential, regardless of how sophisticated AI becomes.
Perhaps the biggest shift, then, is not technological but cultural.
Operations teams are evolving from processors into supervisors. Their expertise is becoming less about manually extracting information and more about validating intelligent workflows, governing AI models and managing exceptions.
The question facing the industry is therefore no longer whether AI can process corporate actions.
In many respects, it already can.
The more important question is whether firms can build sufficient trust — in their data, their governance frameworks, and their operational controls — to allow AI to become a genuine partner in one of asset servicing’s most complex and consequential functions.
For an industry built on accuracy, accountability, and confidence, that may prove the most important corporate action of all.
Few functions within asset servicing remain as dependent on human interpretation as corporate actions. Every dividend, merger, rights issue, tender offer, or bankruptcy begins with an issuer announcement, legal notice, or market communication that must be interpreted before it can be translated into operational instructions. Unlike settlement or reconciliation, corporate actions have resisted full automation because the challenge is not simply processing transactions faster but understanding what they mean.
Issuer announcements arrive in multiple formats across different jurisdictions, while data vendors can interpret the same event differently. Before an instruction reaches a client, operations teams must determine which interpretation is correct.
AI, particularly large language models (LLMs), is beginning to change that. Unlike traditional automation — which relies on predefined rules — generative AI can analyse unstructured documents, extract key information, and summarise complex events in natural language. According to Eurofi, 91 per cent of financial services firms are either assessing AI or already using it in production, with document analysis and knowledge management among the fastest-growing use cases.
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.
Dirk Löscher, head of custody and investor solutions at Clearstream, believes corporate actions are both an ideal use case and one of AI’s greatest challenges.
“Corporate actions are a natural fit for AI because the process still depends heavily on interpreting unstructured issuer announcements, agent emails, and legal documentation,” he says. “At the same time, that is exactly what makes the area challenging: the data is often complex, market-specific, and not always presented in a consistent format.”
While AI can extract and structure information more quickly than traditional workflows, he stresses that “it must operate within strong controls because the economic impact of a wrong interpretation can be material.”
That balance between efficiency and control has become the defining theme of AI adoption across corporate actions.
From proof of concept to production
For several years, AI discussions within financial services centred largely on experimentation. Today, firms are beginning to deploy AI in production, focusing on the most manual parts of the corporate actions lifecycle rather than attempting to automate the entire process.
One of the clearest applications is document interpretation.
Corporate actions teams receive issuer announcements in PDFs, emails, HTML files, and lengthy legal documents that must be converted into structured operational data. Traditionally, this required analysts to manually identify key dates, election options, and event conditions before processing could begin.
According to Adam Cottingham, product manager for corporate actions at SmartStream, AI is proving particularly effective at bridging the gap between unstructured documentation and structured workflows.
“Complex corporate actions often come with detailed offer documentation, which is not written to a standard and is often perceived differently by an individual,” he explains.
“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 automation, AI is becoming an interpretation layer that sits above deterministic processing.
Rules continue to govern entitlement calculations and validations, while AI focuses on understanding documents that previously required manual review.
Löscher describes this as moving workflows “from manual reading, interpretation and re-keying towards assisted extraction, classification, and preparation of structured event data”.
“The objective is not to remove controls, but to reduce manual effort, improve consistency, and increase operational resilience before an event reaches the validation stage,” he informs.
Both firms are now seeing these capabilities move beyond proof of concept.
Clearstream has launched projects to automate asset servicing events, beginning with voluntary repurchase offers before expanding into additional event types.
“This is not AI for its own sake,” notes Löscher. “It is targeted automation of a real operational pain point.”
SmartStream, meanwhile, is using AI to review narrative documentation, summarise event descriptions, and identify conditional language hidden within lengthy issuer announcements. These capabilities remain paired with human validation, allowing operations teams to focus less on locating information and more on verifying it.
AI is also beginning to reshape workflow management. Rather than processing every event in arrival order, firms can prioritise work based on confidence levels and operational risk.
“AI gives the ability to move from deterministic classification to learned, probabilistic classification,” comments Cottingham, “focusing the user on the most important tasks.”
When complexity wins
While AI performs well on routine events, its limitations become clearer as corporate actions grow more complex. Bankruptcies, tender offers, mergers, and restructurings often involve hundreds of pages of legal documentation, multiple election options and evolving terms. Operations teams must not only extract information but interpret legal language, understand market practice, and ensure operational instructions accurately reflect an issuer’s intentions.
Here, AI acts as an assistant rather than a decision-maker.
“Rules-based automation performs well when the input is stable and structured,” observes Löscher. “Corporate actions rarely offer that luxury.”
Instead, he believes AI works best as an interpretation layer, combining document analysis with deterministic validation, and operational review.
Cottingham agrees, arguing that AI’s greatest value lies in accelerating understanding rather than replacing judgement.
“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.”
AI can assemble information from lengthy merger agreements, summarise restructuring documents, compare amended announcements, and identify changes to election deadlines far more quickly than manual review.
“The single biggest benefit,” he adds, “is AI as a reading accelerator.”
However, entitlement calculations, proration methodologies, and election decisions remain deterministic processes that continue to require rules engines and human oversight.
As Löscher concludes: “The right model is not blind automation; it is controlled augmentation.”
The human in the loop
While AI is becoming increasingly capable of interpreting corporate actions, technology providers agree that it should remain accountable to human expertise.
Rather than replacing operations teams, AI is reshaping their role. Historically, specialists spent much of their time reading lengthy documents, comparing vendor feeds, and validating routine events. As AI absorbs these repetitive tasks, operations professionals are increasingly focusing on supervision, exception management, and quality assurance.
Löscher expects that trend to continue.
“In the near term, AI should primarily act as an assistant to operations teams: extracting information, preparing event data, reducing repetitive work, and surfacing exceptions,” he predicts. “Over time, more deterministic parts of the workflow may become highly automated, but decision making in complex or judgement-based scenarios should remain under human oversight.”
Cottingham shares that view, arguing that operations teams are moving “from the doing to supervising”, with expertise shifting towards ambiguous events, complex terms, and fiduciary decisions. While AI can accelerate document analysis, accountability, judgement, and oversight remain inherently human responsibilities.
That human oversight is closely linked to one of the industry’s biggest concerns — explainability.
Unlike traditional rules-based automation, LLMs can reach conclusions without clearly showing how they arrived there. For corporate actions — where every decision must be auditable — that creates an obvious challenge.
“Corporate actions processing takes place in a control-heavy environment, and users must be able to understand why a data point was extracted, where it came from and how it was transformed,” informs Löscher. For Clearstream, trusted AI requires visible lifecycle tracking, validation, regression testing, and clear ownership.
Cottingham agrees that transparency determines whether AI delivers genuine efficiency.
“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,” he says. “Without this, AI merely relocates the manual re-checking rather than removing it.”
As a result, firms are increasingly combining AI with confidence scoring, audit trails, and human approval workflows. Rather than replacing operations professionals, AI is becoming another participant in the workflow — one whose outputs are reviewed, measured, and continuously refined.
The future role of operations is therefore shifting from processing information to governing how intelligent systems interpret it.
Trust by design
As AI moves into production, the conversation is shifting from capability to governance. Financial institutions are increasingly focused not only on what AI can do, but on whether it can be deployed in a way that is transparent, explainable, and accountable.
That discussion has gained momentum with the introduction of the European Union’s AI Act. Although most AI applications used in corporate actions are unlikely to fall within the legislation’s high-risk category, the Act reinforces principles such as human oversight, transparency, and documentation that firms are already building into their AI strategies.
Cottingham believes the regulation will encourage confidence rather than hinder innovation.
“Most asset-servicing AI sits outside the high-risk regime,” he notes. “The EU AI Act is likely to prove less restrictive than feared and more useful than expected.”
He argues that the legislation helps move organisations from questioning whether AI can be trusted to establishing governance that allows firms to “sign off on it”.
Löscher also sees regulation as an enabler rather than an obstacle.
“Regulation will raise the bar, but that is not necessarily a negative,” he states. “In asset servicing, trust is built through governance, controls, transparency, and accountability.”
He expects the AI Act to encourage more disciplined model design, testing, monitoring, and documentation. “That may slow down superficial adoption, but it should accelerate sustainable adoption where AI is deployed with the right safeguards.”
Ultimately, firms recognise that confidence in AI cannot rest solely on accurate outputs. Models must be consistently explainable, auditable, and supported by clear accountability. In that sense, successful AI adoption is becoming as much an exercise in governance as it is in technology.
Better AI starts with better data
While generative AI is transforming how firms process unstructured information, its effectiveness still depends on the quality of the underlying data.
Corporate actions have long been plagued by fragmented information, with issuer announcements arriving in different formats and multiple vendors often interpreting the same event differently. AI can help identify inconsistencies more quickly, but it cannot resolve poor-quality data on its own.
“AI can help identify inconsistencies, missing fields, and unusual data patterns earlier in the lifecycle,” says Löscher. “But it does not magically solve data quality. The value comes from making uncertainty visible: flagging gaps, preparing structured outputs, and routing exceptions to the right operational experts for validation.”
Cottingham agrees, arguing that AI is particularly effective at detecting discrepancies by comparing multiple sources simultaneously. However, he cautions against assuming AI-generated outputs are automatically more accurate.
“AI’s risk in this process is not dirty or incomplete data,” he explains. “It is making incorrect data look clean.”
For both firms, AI reinforces the importance of industry standardisation rather than replacing it. Initiatives such as ISO 20022 and broader efforts to create structured, machine-readable corporate actions data will reduce the amount of interpretation required before information reaches downstream systems.
Löscher believes AI’s greatest contribution may be exposing weaknesses that have always existed. “If the market succeeds in establishing the CSD as the central starting point of a high-quality data flow, AI can help convert that information into structured, machine-readable data at scale.”
Ultimately, AI is proving to be as much a mirror as a solution. Rather than fixing fragmented data, it is highlighting where better standardisation and digitalisation are still needed — making data quality, rather than AI capability, the industry’s biggest determinant of success.
Beyond automation
For decades, the corporate actions industry has pursued greater automation.
Straight-through processing (STP) reduced manual intervention. Standard messaging formats improved interoperability. Workflow technologies increased operational efficiency.
AI represents the next stage of that evolution — but not because it replaces people.
Its significance lies in addressing something previous technologies could not: interpreting information before it becomes structured data.
The technology is already demonstrating measurable value.
It is reducing the time required to analyse complex documentation, highlighting discrepancies between information sources, prioritising operational exceptions, and allowing specialists to concentrate on the events that genuinely require experience and judgement.
At the same time, the industry’s early deployments reveal equally clear boundaries.
Corporate actions remain rooted in legal interpretation, operational accountability, and fiduciary responsibility.
Those are areas where transparency, explainability, and human oversight are likely to remain essential, regardless of how sophisticated AI becomes.
Perhaps the biggest shift, then, is not technological but cultural.
Operations teams are evolving from processors into supervisors. Their expertise is becoming less about manually extracting information and more about validating intelligent workflows, governing AI models and managing exceptions.
The question facing the industry is therefore no longer whether AI can process corporate actions.
In many respects, it already can.
The more important question is whether firms can build sufficient trust — in their data, their governance frameworks, and their operational controls — to allow AI to become a genuine partner in one of asset servicing’s most complex and consequential functions.
For an industry built on accuracy, accountability, and confidence, that may prove the most important corporate action of all.
NO FEE, NO RISK
100% ON RETURNS If you invest in only one asset servicing news source this year, make sure it is your free subscription to Asset Servicing Times
100% ON RETURNS If you invest in only one asset servicing news source this year, make sure it is your free subscription to Asset Servicing Times
