IBM's deal to buy Promontory Financial Group portends a dramatic change in the roles computers and humans play in regulatory compliance — and perhaps banking generally.

If the deal announced on Thursday goes through, the 600 consultants at Promontory will have the task of teaching Watson, IBM's massive artificial intelligence engine, how to handle risk management and compliance chores for financial services firms. Such automation may not make bank compliance officers obsolete, but it could mean that far fewer of them will be needed in the future, and that their time will be devoted to higher-level tasks.

"I immediately thought of all the lawyers and former government regulators that work at Promontory being replaced by computers," said Ryan Gilbert, partner at Propel Venture Partners. "If truly the purpose of this acquisition is to take the human knowledge and effectively store it in AI or Watson, it will have a huge effect on this industry."

AI is a hot topic for banking in general. Banks are thinking about or using artificial intelligence technology and its sidekick, robotic process automation, throughout their organizations: in lending, in detecting fraud, in operations, in human resources, account opening, complex payments and elsewhere. Essentially any place humans are doing mundane, routine work, someone's thinking about deploying AI to do it instead.

Eugene Ludwig, the founder and CEO of Promontory, said that just as Watson assists with medical diagnoses without putting doctors out of work, neither would it fully automate the work of bankers or compliance officers.

"There will always be a need for hands-on, bespoke solutions to individual problems," said Ludwig, a former comptroller of the currency. "For us, we will continue to do the hands-on work that we do today."

IBM's plan for Watson to absorb the collective wisdom of Promontory's brain trust is part of a recent phenomenon known as regtech. Like fintech, this field attempts to digitize outdated analog processes, replacing stacks of paper and hours spent looking things up and cross-checking with fields of data and fast answers.

"Many compliance vendors are looking to beef up their products with inorganic intelligence technologies, and they're increasing their efforts and footprints in the financial services space," said David Weiss, a senior analyst at Aite Group.

There is an effort to make digital advice technology more intelligent and into a tool designed to assist advisers rather than compete with them. (Bloomberg News)
There is an effort to make digital advice technology more intelligent and into a tool designed to assist advisers rather than compete with them. (Bloomberg News)

In February, for example, Nasdaq formed an alliance with a cognitive computing company called Digital Reasoning to provide surveillance technology to buy-side firms, brokers, regulators and exchanges. The following month, Credit Suisse and Palantir Technologies formed a venture called Signac that will use artificial intelligence to try to catch rogue employees before they can breach conduct rules.

Aite estimates financial services spending on compliance tech will grow more than 10% over the next several years.

"It's one of the very few areas of such growth," Weiss said. "Financial services compliance has reached peak human conditions — you can't just throw bodies at the problems. So this is a tremendous market for IBM to get into and they know it."

Analysts at the consulting firm Bain & Co. estimate that governance, risk and compliance costs account for 15% to 20% of the "run the bank" cost at most large banks and 40% of the costs of innovation projects at banks.

"The complexity of regulation has grown to a proportion that is untenable to most law firms," said Brian Roemmele, a researcher and analyst and co-host of the "Around the Coin" podcast.

More than 100 startups already provide regulatory technology solutions, according to the law firm White & Case. Gilbert said he expects IBM's acquisition to spur further VC investments in such companies, both early and later stage.

IBM has been trying for years to sell Watson — the massive, "Jeopardy!"-winning software set that takes in documents and data and learns to do tasks formerly done by humans — to banks. A few, such as ANZ in Australia, use it to help financial advisors understand their clients and make product recommendations. In the United States, USAA uses Watson to help veterans transition to civilian life, by answering their questions.

"It's nice to win 'Jeopardy!' but that's not a revenue source," said Ohad Samet, co-founder and CEO of TrueAccord, a startup applying machine learning and analytics to debt collection. "This shows a lot of foresight on IBM's side," he said of the deal to buy Promontory.

Some question whether IBM and Promontory could truly create a Skynet for bank compliance.

"We're going to have to see how this marriage works out and to what extent artificial intelligence can functionally contribute to compliance," said Brian Knight, a senior research fellow in the financial markets working group with the Mercatus Center at George Mason University. "AI can be valuable in certain areas, but what about the types of nuances humans are better at?"

Much will depend on how IBM and Promontory work together. For example, will Promontory largely continue as is, but with the capability to tap into Watson if needed? Or would Watson have access to Promontory's bank clients' files and review them and alert the bank with a message when it notices a compliance concern?

"We're going to have to see how this marriage works out and to what extent artificial intelligence can functionally contribute to compliance."

"There's different ways this could go; whether a people-first model or an AI-first model," he said. "I suspect it will probably be a mixture."

Putting more data in various data stores always raises concerns about cybersecurity and privacy, noted Daniel Kimerling, head of API banking, open platform, and R&D at Silicon Valley Bank.

"In general, the question is, do we want to trust machines?" he said. "And how do you verify the machines — not only that the code accurately models regulatory requirements but whether the systems are properly implemented inside the financial institution?"

Weiss at Aite, who has been an articulate evangelist for the recent renaissance of artificial intelligence, nevertheless has doubts about whether Watson could solve all of banks' compliance problems.

"IBM Watson is very capable, but the devil is in the training — by Promontory — and monitoring by compliance pros post-implementation," he said. "As a bionic solution it could work very well. As a pure compliance replacement it would eventually fail."

What would a Promontory-led, IBM Watson-driven compliance department look like?

"If these robo banks will be able to ingest all the good knowledge that consultants at Promontory have, and make decisions without emotion, we'll probably see better decisions being made," Gilbert suggested. "Hopefully Watson the banker will be able to do away with many of the ethically and emotionally driven problems we've seen in banking."

Kimerling also envisions much-needed change in compliance groups.

"People are not very effective," he said. "Machines are somewhat effective. People plus machines are extremely effective."

Kimerling said he envisions a future where inside banks "machines will take care of the 80% of cases that take 20% of the work, and people will concentrate their energy on the 20% of cases that take 80% of the work."

Fewer people will be employed in compliance departments, but those who remain will be able to "concentrate on edge cases and complex gray areas," he said.

Instead of thousands of compliance officers, banks might instead have hundreds, who will be "tasked with really thinking about places where the rules don't apply in a neat and consistent fashion," Kimerling said. "In some ways this will force regulators to think about how you write regulations so it can be easily abstracted into code."

Realization of the regtech vision may require a culture shift in Washington, though.

"I'd love for regulators to accept data over API," Kimerling said. "But try to get a regulator to understand what API means, let alone accept data over that. It's complicated."

Gilbert also wondered how the banking agencies will adapt.

"How will regulators respond," he asked, "when financial institutions say they're basing their decisions on Watson?"

Mary Wisniewski, Bryan Yurcan, Tanaya Macheel, Rob Blackwell and Marc Hochstein contributed to this story.