- Citigroup is partnering with artificial-intelligence-powered fraud-detection startup Feedzai to bolster the bank's $8.5 billion Treasury and Trade Solutions business.
- As a first step, the companies are collaborating on a tool to better protect the $4 trillion in payments the bank's thousands of corporate and public-sector customers send every day.
- That fast-paced business is increasingly under threat from cyber crime.
- Employing Feedzai's machine-learning capabilities, Citi is souping up a tool that alerts customers to suspicious or aberrant payments.
- Citi expects this will cut down on costly and time-consuming false positives and make the tool applicable to more clients.
Citigroup has partnered with an artificial-intelligence-powered fraud-detection startup to help protect the trillions of dollars in payments the bank's customers send every day.
Citi on Wednesday announced a broad partnership with Feedzai, a fintech company the bank has previously invested in, and their initial collaboration involves adding firepower to a Citi tool that alerts companies to aberrant or suspicious payments.
Through its crucial Treasury and Trade Solutions business, which pulled in $8.5 billion in revenue last year, Citi processes $4 trillion in payments each day for thousands of corporations and public institutions across the world.
It's a rapidly evolving business that's increasingly coming under threat from sophisticated cyber crooks — a record 78% of organizations fell victim to payment fraud last year, according to a 2018 survey from the Association for Financial Professionals.
To safeguard its customers, Citi late last year piloted a tool with a limited number of customers to detect unusual or outlier payments — say, a $5 million transaction sent at an odd hour of the night or to an unexpected country.
That aberrant payment could have a number of explanations, some innocent — an employee made an error and fat-fingered a transaction — and some nefarious, such as an unfolding cyber attack.
"The underlying reasons could be many, but what the tool does is it uses our own channels to identify the payment and go back to our customers and say, 'Is this really you? Is this the payment you intended to do?" Manish Kohli, global head of payments and receivables in Citi's Treasury and Trade Solutions division, told Business Insider.
The solution, called Citi Payment Outlier Detection, was popular with customers, Kohli said, but it wasn't perfect.
Alerts like this, based on broadly applied rules, don't account for the nuances and different behaviors of the thousands of organizations Citi manages money for. That can result in false positives that gum up a process that moves at a breakneck pace and is only getting faster amid the adoption of digital and automated-payment technology.
It can also numb the customer to the alerts, leading to them letting their guard down — the "crying wolf" dilemma, as Kohli put it.
"A broad-based, rules-based approach is not the right way to address and find outlier transactions," Kohli said. "If you have a significant number of false positives, that actually takes away the whole efficiency and value of a product and solution like this."
That's where Feedzai and its machine-learning and AI capabilities come in.
Citi Ventures invested in the startup, which was founded in 2009 and has raised $82 million, back in 2016, and the bank recognized that Feedzai's capabilities could solve a variety of problems and headaches.
An obvious place to start was enhancing Citi's payment-outlier-alert system.
Harnessing the power of machine learning, the alert system could adapt to the behaviors of individual customers, creating a profile for the client as a payment pattern took shape.
"We recognize with institutional clients, every client is different. And their payment pattern is different," Kohli said.
The souped up, AI-powered version of the tool should detect suspicious behavior with more precision and cut back on false positives, keeping business humming along at a fast clip until an actual threat or mistake is unearthed.
Payments are assessed a risk score, and the threshold that triggers an alert from Citi can be tailored to the customer's needs.
Kohli said this would allow Citi to roll out the Citi Payment Outlier Detection system to a broader set of companies starting in early 2019, when the enhanced version of the feature is expected to be released.
This is only the first phase of the partnership, and Citi will look to apply Feedzai's capabilities to other products and services in its business as well, Kohli said.