The rapid advance of generative AI content creation tools such as voice cloning, video and image generation, and chatbots have already resulted in new forms of fraud and financial crime: recall the unfortunate worker in Hong Kong who was recently tricked into transferring $25 million to fraudsters impersonating his CFO.
But even as the threats multiply thanks to gen AI — and the losses pile up, with an estimated $485.6 billion lost to fraud globally in 2024, according to research from Nasdaq — technologists are working on new ways to use AI and machine learning (ML) to help apprehend and deter those responsible.
Take the San Francisco-headquartered startup Hummingbird, which launched in 2016 specifically to offer easy-to-use, secure tools for in-house crime investigators and compliance teams at large financial institutions and payments providers (among its customers are payments platform Stripe, bespoke marketplace Etsy, the sports betting platform DraftKings, FirstBank Puerto Rico, and others).
Today, in a VentureBeat exclusive, Hummingbird is announcing a new software automation product for customers, aimed at eliminating the manual, repetitive tasks that bog down investigators and compliance teams at large financial institutions.
VB Event
The AI Impact Tour – NYC
We’ll be in New York on February 29 in partnership with Microsoft to discuss how to balance risks and rewards of AI applications. Request an invite to the exclusive event below.
Request an invite
“The goal is really bite-size automation recipes that help with financial crime investigations,” said Joe Robinson, CEO of Hummingbird, in a video call interview with VentureBeat. “And because this work is so regulated, any applications of AI or logic or algorithms in general need to be relatively transparent, very auditable, and process-focused with a human in the loop. Those are the key characteristics of what we’re doing.”
The murky world of fi-crime investigations
Investigating financial crime may seem to those outside the industry like something from an episode of “Law and Order,” filled with intrigue, drama, and suspense.
And it certainly can be. But according to Robinson, the meat of the work is in combing through vast volumes of financial transaction data to find anomalies and determine if they are worth a closer look, as well as trace where the money is coming from and going to — is it being laundered, or derived from criminal activity?
For banks and companies that accept payments, offer payment platforms, account solutions or wallets, the move in recent years has been to hire teams of dedicated, in-house investigators to do this job regularly, then call in outside authorities when proof of a crime has been assembled.
“When you’re a financial crime investigator, you’re generally looking at the activity of one of your customers and trying to suss out: ‘Is the activity suspicious or not?’” Robinson explained.
“For example, if I was at a payment card company, I might be looking at 15 to 30 different data sources that represent a specific customer’s activity. It’s slow, pretty tedious work. Data fragmentation is one of the major issues. And then later on, you might be audited for why you made certain decisions related to that account and how you handled it, so your decision-making process needs to be transparent. All those things make this a unique market for applications of AI, of machine learning, of different algorithms.”
Ideally, Robinson said that the Hummingbird Automations platform will help assemble information on fairly routine financial crimes such as tax evasion while freeing up human investigators to focus on harder, less automatically detectable issues such as human trafficking.
Yet even in the former case, the platform only presents the information it has gathered to a human officer — it’s up to them to review it for accuracy and then ultimately send it off as a report to their supervisor or outside law enforcement, keeping them in the loop as the expert and controller.
Pre-built ‘recipes’ for automating investigation drudgery while maintaining high accuracy
Hummingbird Automations is here to help investigators through the course of all this with a library of pre-built “recipes” or workflows, addressing key compliance activities shared across large financial institutions. Among these are:
- Know Your Customer / Business (KYC/B): Track, monitor, and trigger reviews of your customers, with support for ongoing and periodic monitoring.
- Quality Assurance (QA): Schedule QA reviews on a sample of recent cases to verify the quality of completed work.
- Case Preparation: Accelerate investigations by pre-populating case information and findings.
- Case Monitoring: Automatically notify users of cases that require their attention, such as those with an upcoming deadline or a rejected filing.
- Case Management: Automate administrative tasks, such creating, assigning, and updating cases.
- Activity Digests: Improve oversight with regularly scheduled reports of case activity.
These features are not just about saving time; they’re about enhancing the effectiveness and oversight of financial crime investigations, making compliance programs more efficient and less prone to risk.
“AI might be used to summarize some of the information available to the investigator and help them interpret what’s going on,” Robinson explained. “I think in contrast to other markets, AI is really good at spotting trends and spotting anomalies across lots of data and documentation, so that’s where we’re applying it at Hummingbird.”
Early positive feedback
Robinson told VentureBeat that the Automations platform has been in the works for nearly a year, and has been tested with a small group of about five customers in a closed beta release since spring 2023.
Hummingbird provided the following statement from an anonymous “buy now, pay later” (BNPL) provider, who participated in the beta: “Hummingbird Automations has made a huge impact on my team and our work. We’re now able to easily automate notifications and reporting, allowing us to increase oversight and streamline our workflow.”
As financial institutions grapple with the dual challenges of regulatory complexity and the need for cost-effective, scalable investigation tools, Automations by Hummingbird offers new AI-powered efficiencies, while maintaining high levels of transparency, human oversight, and reporting for compliance.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.
