Corporate actions—such as dividends, stock splits, rights issues, and mergers and acquisitions—affect every security, and require the distribution of information to investors, the capture of those investors’ decisions, and the “pricing-in” of those actions to a security’s price. It’s a long-standing and largely manual process that has changed little over the years. And as securities become more sophisticated, that process is only becoming more complicated.

Now, tech giant Google is aiming to simplify that process, using artificial intelligence (AI) tools originally developed for use in text extraction and call centers.

“There are people on the buy side who want to trade quickly on corporate actions. But it’s a significant challenge with lots of bodies in the middle,” says Philip Moyer, vice president of strategic industries (comprising healthcare, life sciences, retail, consumer products, telecoms, media and entertainment, and gaming, as well as financial services) at Google Cloud.

Google has many other business areas that are far more profitable than corporate actions. Once they realize how hard this is, and how much it costs, they may back away. But if they partner with someone who knows what they’re doing, Google won’t have to do so much groundwork.

Virginie O’Shea, Firebrand Research

Most financial institutions get their corporate action data from multiple vendors, and sometimes the information contained in those feeds don’t match up, thus the user has to manually decide which feed is best. Moyers says that feeds often disagree on details because each vendor manually extracts and interprets the data from an issuer’s original announcement, and also because each vendor uses their own taxonomy.

“So financial institutions are left to manually process and compare these. And when making decisions for your assets, extraordinary amounts of exceptions arise where people need to get involved,” he says. “Firms can employ hundreds of people to do this.”

Barry Raskin, managing director at market data and management consulting firm Jordan & Jordan, says he’s not surprised that someone would seek to apply AI to corporate actions, agreeing that it could significantly reduce the amount of manpower employed within financial firms to handle corporate actions processing.

“One major US investment bank used to have a whole floor of people in Tampa, Florida, just processing corporate actions. And firms probably still have significant operations handling this, because it’s critical,” he says.

Listen and learn

Moyer envisages using a combination of two existing Google solutions to address corporate actions—its DocAI tool for text extraction, and its Contact Center AI solution, which listens to chat sessions in customer call centers to determine what issues are being handled, and whether issues need to be routed to a live agent.

Companies have been using text extraction at various levels to understand corporate actions for many years—at least for a decade, when Moyer was CEO and president of financial and regulatory data and compliance reporting tools provider Edgar Online.

“Even at Edgar, we were parsing documents, doing text extraction using optical recognition to pick out words,” Moyer says, though this was primitive compared to what Google can do now,” he says. “DocAI looks at documents as an entire document, so we retain complete fidelity of the document and keep the overall context of it, so that when we do extraction, if we define the term ‘effective holder’ in one place, we can associate that with other uses elsewhere in the document.” 

In addition, clients can use Google’s Knowledge Graph to enrich a corporate action by capturing and pulling together information not contained in the document itself—for example, to add information about where a company is headquartered, or what currency the corporate action is denominated in.

Marry that with the Contact Center AI tool, which continuously learns from itself and becomes more intelligent. “You can imagine how the AI we use in call centers could, for example, make a decision about a dividend payment,” Moyer says.

Automation vs. intervention

Virginie O’Shea, founder and CEO of Firebrand Research, welcomes the potential of new entrants using new technologies to address the challenges of corporate actions, but warns that it won’t be easy for newcomers.

“There are very few vendors that have lasted in this space. Twenty years ago, there were a lot, but now it’s just a handful. And the reason is because it’s very hard to automate. And if you don’t understand the process, then you won’t make a difference,” O’Shea says. “In terms of financial institutions, I think they are all struggling with the same problem. But they have an aging workforce doing this, and young entrants to the workforce don’t want to handle it. So firms need to automate.”

One barrier to this is that there is no incentive for the issuers themselves to simplify the process. Corporate actions themselves are complicated, as is the process by which they are managed, with companies employing “a chorus of intermediaries who deal with investors, custodians and brokers,” O’Shea says. “They don’t want to spend money to change the process.”

In the absence of change at the issuer level, AI presents a solution to simplify that complexity. And while AI can be a driver of greater automation because it’s good at spotting patterns, which makes it good for handling unstructured data—and much of the underlying information associated with corporate actions is usually in unstructured form—O’Shea warns that AI models take a long time and a lot of data to train.

And even if you have the data and AI tools, corporate actions require a lot of specialist expertise—which may require investment in human talent or partnerships with existing providers.

“Google has many other business areas that are far more profitable than corporate actions. Once they realize how hard this is, and how much it costs, they may back away. But if they partner with someone who knows what they’re doing, Google won’t have to do so much groundwork,” she says.

Raskin spent three decades at Six Financial Information and predecessor Telekurs Financial, one of the largest providers of corporate actions data, so he knows a thing or two about corporate actions. And while he’s positive about the potential for AI, he also warns that applying it to corporate actions won’t be easy.

“Back at Six, we did a lot of automation. You can automate and use AI to do just about anything with corporate actions on plain vanilla securities. But the problem is complex corporate actions and complex securities. Managing those is quite difficult, and there are new instructions and new corporate actions every day,” he says. “For example, the underlying could be a convertible bond, or there could be a complex election involved in the corporate action … so there will always be a need for some level of human intervention.”

But Google isn’t deterred: While there’s no specific data for a formal launch of its services for corporate actions, Moyer says Google is having active conversations with a lot of data providers, financial firms, and custody agents about how its tools could be used in this space, and will release a library of corporate action types once the vendor feels comfortable that it has processed enough documents.

He emphasizes that Google will not become a corporate actions data provider, but will rather provide tools that make providing and consuming corporate actions data easier for all participants in this space—similar to how its Lending DocAI tool helps automate mortgage processing.

“That doesn’t mean we’re becoming a mortgage processor, and, likewise, here we are not going into the business of competing with existing corporate actions providers. If vendors and custody agents want to use these tools, we want to work with them,” he says.

And if its corporate actions efforts prove successful, the vendor plans to apply its tools to other important areas of the financial markets that still require human intervention, such as Know Your Customer and anti-money laundering requirements. For example, clients can use DocAI for parsing and extracting data from documents, then verifying information—such as fundamental data about a security, client or counterparty—using the Google Knowledge Graph.