Home›Companies›Finic

Automate fraud investigations with AI

Finic helps fintechs automate fraud and dispute investigations with AI agents that automatically follows SOPs and writes comprehensive case notes. With Finic, fintechs can: - 10x fraud ops capacity without increasing headcount - Train agents on new fraud typologies and instantly scale to thousands per day - Reduce time-to-resolution from hours to under 5 minutes

Finic
Founded:2024
Team Size:2
Location:San Francisco
Group Partner:Dalton Caldwell

Active Founders

Jason Fan, Founder

Ex security & product @ Robinhood. University of Toronto Computer Science.

Jason Fan
Jason Fan
Finic

Ayan Bandyopadhyay, Founder

Founder @ Finic (YC W23). Caltech class of 2020, former software engineer at Robinhood.

Ayan Bandyopadhyay
Ayan Bandyopadhyay
Finic

Company Launches

We’re Jason and Ayan, the founders of Finic. We help fintechs automate fraud investigations with AI agents.

tl;dr

Finic lets fraud ops teams at financial institutions scale up their capacity by 10x without increasing headcount. We use intelligent automation to analyze structured data (SQL queries, risk scores) and unstructured data (customer support tickets, adverse media), allowing us to triage tickets and generate case notes twenty times faster and at 80% lower cost compared to tier 1 fraud analysts.


—


At Robinhood, we led the identity team. One of our key mandates was to prevent, detect, and investigate fraud, with a specific focus on identity fraud. We were able to reduce account takeover (ATO) fraud losses by 95%, but it took us 10 months and a lot of new hires to get there, largely due to bottlenecks in the human review process. We tried automating fraud SOPs, but weren’t able to arrive a compelling solution because at the time there wasn’t a way to automate the biggest bottle neck to fraud reviews - subjective reasoning.

Now, we’re building Finic to be the product we wish we had, with AI that now makes this possible.

Why is this important?

Consumer financial fraud losses crossed $10bn for the first time in 2023, and is likely to keep growing as new technology also enables new attack vectors, most recently with synthetic identities created by GenAI.

The way financial institutions prevent fraud is usually split into three parts:

  • Prevention: Find and closing loopholes that enable fraud in the first place.
  • Detection: using ML or heuristic models (aka “rules engine”) to detect unusual activity.
  • Response: Investigating cases of potential fraud, deciding on a course of action, and, in some cases, writing reports detailing the investigation.

Prevention is typically achieved through product improvements, and the market for data products that help with fraud detection is mature. However, the way that companies respond to fraud is still mostly done by human review by a combination of highly skilled onshore and low-skilled offshore analysts.

Problem

Using humans as the front line of fraud investigations is not ideal for several reasons:

  • Humans make mistakes. For CS tickets, mistakes mean frustrated customers. For fraud cases, mistakes can result in thousands or millions in losses and compliance headaches that last years.
  • Humans take a long time to train. Fraudsters change tactics frequently. By the time a team of analysts is fully trained on a new fraud typology, fraudsters have already moved on to other strategies.
  • Humans are expensive to scale. The average offshore agent costs $43,000 per year and has a throughput of about 250 cases per month. This means it costs companies $14 per case to investigate fraud, and that’s not even factoring in the period it takes to onboard and train new agents, which is a constant problem since turnover is so high.

Financial institutions aggressively overhire when fraud is bad, yet still find themselves unable to respond quickly to new fraud typologies, leading to millions in preventable losses per quarter. As a result, every fraud ops team I’ve worked with is either understaffed and overworked, or overstaffed and just waiting to be laid off.

Solution

Finic allows fraud ops teams to build custom AI agents that execute SOPs autonomously, improving on the work of junior fraud analysts and allowing the team to scale capacity by 100x with the click of a button.

This gives fraud ops teams several superpowers:

  • Accuracy and explainability: Finic agents document each step of their workflow, making it easy for senior analysts to audit, which means fewer mistakes. Case notes written by human analysts take much longer and are often inconsistent.
  • React faster to new fraud vectors: Finic agents can be created, trained, and scaled to 1000s of cases per day in a matter of hours, something that would be impossible for a team of human analysts.
  • Scale capacity instantly: Fraud ops can now scale capacity as easily as engineering can scale web services. Finic agents complete investigations in 1-5 minutes, compared to 1-2 hours for human analysts.

The Finic platform has three parts, all of which are built on top of a multimodal reasoning engine.

SOPs

Each Finic agent is defined with a SOP, or list of actions they need to take in order to investigate a ticket. This can include data retrieval steps to build a profile of the customer’s activity, analysis steps to reason about whether fraud is taking place (and if so, what kind), as well as actions the agent can take once a decision is made, like escalating tickets for human review.

Customers can create an SOP on Finic from scratch or import an existing document to automatically generate each action based on existing SOPs.

Tools

Fraud analysts need many tools to conduct an investigation, from internal dashboards to SQL tables to vendor data.

Finic uses intelligent automation to connect to any data source and use it the same way a human analyst would: through a browser.

Rather than waiting for the engineering team to build new APIs, all it takes to get started with Finic is a list of credentials for the tools they use, drastically reducing the time it takes to get to ROI.

Audit Logs

Whether the investigator is an AI agent or a human analyst, keeping a trail of why decisions were made is just as important as making the right decision. In fact, several regulations mandate that financial institutions keep logs of not just transactions but also the reasons why transactions or accounts were flagged as fraudulent.

Finic agents “think out loud” and document each step in the SOP, including what data sources they used, how they arrived at their conclusions, and what steps they took afterward. This makes it easy for senior analysts to review agent activity to ensure compliance and is usually more comprehensive than the reports generated by junior analysts.

Learn more

Here’s a list of some of the types of fraud we’re helping customers remediate:

  • Identity Fraud: Reviewing KYC tickets to determine if new signups are using stolen/synthetic identities, or if the account is being used as a proxy by a malicious third party.
  • Friendly Fraud: Reviewing account activity and dispute packages from third parties to determine if customer-reported issues like credit card chargebacks are legitimate.
  • Account Takeovers (ATO): Reviewing accounts flagged with unusual login activity, and looking at device fingerprint, IP address, transaction activity and other signals to determine if the account has been compromised.

If you’re a fintech looking for ways to 10x the capacity of your fraud ops team without increasing headcount, email me at jason@finic.ai, and I’d be happy to show you a demo.

Other Company Launches

🌱 Dealwise: an AI powered investment bank for startups

Get acquisition offers in just 3 months. No exclusivity, cancel anytime
Read Launch ›

YC Sign Photo

YC Sign Photo