We recently announced Risk AI, the first SMB-specific autonomous agent that acts and makes decisions like a risk analyst. We’ve already received an overwhelming amount of positive feedback from enterprises and start-ups. 

Today, we’re announcing underwriting capabilities within Risk AI. Now, teams can use the agent as an assistant for core underwriting responsibilities, such as merchant website analysis, EIN/TIN verification, OFAC/AML screening, and much more.

Read on to learn more, or contact us if you’d like to get started today.

‍

The problem 

We’re witnessing record growth in small business formation, but the tools to onboard, underwrite and monitor these SMBs are lightyears behind the tools available in consumer finance. As a result, underwriters must process an increasing number of applications with limited tools, headcount, and budgets.

For each SMB application, underwriters have a checklist of data to review before deciding whether to onboard the applicant. Most of this data – online review text, web presence metadata, etc. – is unstructured and stored in different places. Static risk rules and models cannot be applied to this data because they primarily rely on structured signals. 

Take Secretary of State (SOS) data, for instance. Each state has its own SOS database with business registration information, and some states don’t make this information publicly available. Underwriters must get access to and search each SOS database for an SMB applicant’s business registration details, and they’ll often conduct additional manual online searches to fill in any data gaps. This kind of process can delay SMB onboarding timelines by 100 days and creates opportunities for subjective, inconsistent risk decisions.

‍

How do LLMs fit in?

Large language models (LLMs) are a better fit for unstructured data than conventional methods of static risk rules and models. LLMs address many of the above challenges by systematically parsing unstructured data to capture its semantic meaning. Importantly, they do this without losing important context or inviting subjectivity into the analysis. LLMs also excel at solving probabilistic problems such as underwriting, where an entity must make context-aware decisions using a risk-based approach.

We’ve already successfully implemented LLM-powered risk tools to automate risk management for large enterprises, and recently launched Risk AI’s inaugural use case focused on manual review automation. Customers kept mentioning underwriting pain points to us, so we asked ourselves: can we use AI agents to streamline underwriting and maintain a high bar for accuracy?

‍

How it works

Risk AI’s underwriting component is a 24/7, Slack-based conversational AI agent that automates underwriting decisioning. It reduces underwriting times from 20+ minutes per merchant to 1-2 minutes per merchant. This translates into significant time savings for fast-growing teams.

With a simple prompt - such as “Please do an underwriting review of all merchants onboarded in the last 24 hours” - the agent will identify merchants that fit this criteria and conduct analysis across key attributes:

• Merchant website professionalism, based on our proprietary model analyzing for the presence of certain pages, presence of prohibited/restricted terms, accepted payment methods, merchant type, and more
• Adverse media screening on businesses and associated personnel
• Business registration details (e.g., registration status, business age, and other details pulled from SOS databases)
• EIN/TIN Verification
• Review platform reputational data (e.g., Facebook, Google, Yelp)
• OFAC/AML screening

‍

After conducting this analysis, Risk AI will add a note for each merchant with its recommended risk decision: approve, decline, or forward to manual review. These notes are included in the summary message shared in Slack.

Customers can easily create their own SOP and specify which of the available capabilities they’d like Risk AI to access (e.g., online research, KYB, merchant reachout, etc. ). This customization preserves customer privacy while maintaining a streamlined workflow.

‍

Want to learn more?

Underwriting is just the beginning of what our autonomous agent is capable of. 

‍Reach out if you’d like to learn more about Risk AI’s underwriting capabilities, or if you have use cases you’d like us to automate with the agent.