Roe AI
San Mateo, US · 6 known investors
Roe AI builds AI agents that investigate fraud, dispute, AML, KYC/KYB, and sanctions cases end-to-end, reading customer, transaction, and case data where it lives and producing cited, audit-ready dispositions and SAR narratives. It serves compliance and fraud operations teams at card networks, BNPL providers, marketplaces, and crypto companies, operating as a browser-based layer on top of existing tools with single-tenant VPC deployment.
Also known as Roe · ROE · Roe AI
Founders & leadership
Investors · 6
Also in the syndicate · 4
Funding
SEC filings, press & company announcements- Undisclosed amountSeed FundingAug 2024Source ↗
Source: company announcements and press reports — follow each round's link for the claim.
Company profile
researched Aug 2026Roe AI (branded "Roe", also listed as ROE on Y Combinator) develops AI agents that carry fraud, dispute, AML, and KYB cases through to a documented disposition rather than only summarizing alerts. The system reads structured and unstructured evidence, follows a customer's own standard operating procedures, and produces dispositions and SAR-ready narratives in which each claim cites the field, document, or query that produced it. Outputs are posted back into the customer's existing case management, CRM, operations, and customer-support tooling.
The company positions the product as a layer on top of an institution's existing stack rather than a replacement: it connects to risk data providers (sanctions, PEP, adverse media, KYC/KYB, business filings, credit signals) and to data warehouses holding customer, transaction, and case records, and it operates a browser agent for legacy consoles, partner portals, and internal tools that lack APIs. Deployments ship with typologies, policies, and audit formatting matched to the customer's program, and cover card issuers and processors, buy-now-pay-later providers, marketplaces, sportsbooks and prediction markets, insurance carriers and MGAs, and banks and sponsor banks.
Roe AI was founded in 2023 and went through Y Combinator's Winter 2024 batch. At launch its product was described as a next-generation data warehouse that applied multimodal AI agents to unstructured data — documents, images, websites, video, audio — queryable through SQL combined with natural language, with connectors to S3, Snowflake, and Databricks; the company has since focused on agentic risk and compliance investigation.
Founding story
Roe AI was founded in 2023 by Richard Meng, who studied CS and statistics at UC Berkeley (2014-18), led knowledge graph and generative skill assessment work at LinkedIn, and served as tech lead of the generative AI initiative at Snowflake, where he built an AI data copilot served to Fortune 2000 customers. A co-founder identified in launch materials as Jason, also a UC Berkeley graduate (CS and applied mathematics), had previously worked on ML infrastructure at Meta and led infrastructure and DevOps teams at Robinhood and Retool. The pair's original thesis was that conventional data warehouses served only the roughly 20% of enterprise data that is structured, leaving documents, images, video, and audio largely unused in business decisions.
Business model
Roe AI sells software to enterprise risk and compliance functions, deployed as single-tenant infrastructure inside the customer's own cloud environment (AWS, GCP, or Azure) or on-premises, with customer-managed encryption keys, configurable retention, and no training on customer data. The company publishes cost-per-case comparisons against offshore fraud operations, indicating the product is positioned as a substitute for analyst headcount and outsourced review capacity.
Traction
The company reports handling more than 20,000 daily reviews and serving customers with an aggregate market capitalization above $400 billion. Aggregated figures from live deployments are stated as over 90% day-one accuracy on production cases benchmarked against analyst-graded ground truth, 70% cost reduction versus offshore fraud operations (based on 2025 customer cost-per-case studies), 91% faster time-to-disposition, and a fastest time to first production decision of one week across deployments completed in 2025. At the time of its seed round, its platform was in use at Arc, Revere, and Crossmint and in pilot with Fortune 2000 companies. Y Combinator lists a team size of 9; the company site lists four open roles. Coverage has appeared in The Nilson Report and CUInsight.
Latest developments
The company publishes customer and engineering posts, including an engineering piece on its AI-native development practices (January 2026), a piece on limits of KYC co-authored with an ex-Head of Risk Product at Block (April 2026), and a customer story on Dutchie's compliance program (July 2026). It reports deployments completed in 2025 and cost-per-case studies from 2025, and is hiring for four roles across revenue and engineering in San Mateo.
▸Full profile — market position, technology, go-to-market, geography, history
Market position
Roe AI frames the incumbent alternatives as hiring or outsourcing analyst capacity versus relying on rules and models, and differentiates against other AI agents that summarize alerts while leaving the decision with the analyst. It is categorized by Y Combinator under fintech, generative AI, compliance, and regtech.
The company contrasts its approach with other AI agents on five axes: investigating to a defensible disposition rather than summarizing; citing the field, document, or query behind every claim instead of black-box output; multimodal reasoning across warehouse SQL and unstructured artifacts rather than a single modality; continuous re-tuning from QA feedback rather than static models requiring manual retraining; and single-tenant VPC deployment with customer-managed keys and per-case audit logs rather than a shared SaaS data plane.
Technology
The platform runs multimodal AI agents that reason jointly over SQL queries against a customer's warehouse and over unstructured evidence such as PDFs, images, websites, social profiles, and call recordings. Every prompt, source, and decision is logged immutably and can be exported to a GRC system; SOPs and prompts are git-tracked so changes can be rolled back and historical cases replayed against either version. QA feedback is used to re-tune policy alignment continuously. A browser agent extends coverage to web applications without APIs. The company states SOC 2 Type II, HIPAA, and GDPR attestations, role-based access with SSO and SCIM, and model documentation, validation evidence, and challenger-model support aligned to SR 11-7-style model risk governance.
Go-to-market
The company markets primarily through demo requests on its website, published customer stories and bylined articles, and a trust portal documenting security attestations. It cites named customer references from Dutchie, Affirm, and Crossmint, and makes benchmark results available under MNDA. An advisory group of risk, compliance, and data executives from Snowflake, Block, Intuit, Afterpay, Streamlit, Zeni, and Taekus is described as informing product and policy decisions.
Compliance, fraud, and risk operations teams at heavily examined financial institutions, including card issuers and processors, card networks, buy-now-pay-later providers, marketplaces, sportsbooks and prediction markets, insurance carriers and MGAs, banks and sponsor banks, and crypto-native compliance teams. Named users include Dutchie, Affirm, Crossmint, Arc, and Revere, plus pilots with Fortune 2000 companies.
Geography
Headquarters are listed as San Francisco on the company's about page and Y Combinator profile, while the founding team and all open roles are described as based in San Mateo, California.
History
The company was founded in 2023 and participated in Y Combinator's Winter 2024 batch, launching as a data warehouse for unstructured and structured data powered by generative AI, with SQL-plus-natural-language querying and connectors to sources including S3, Snowflake, and Databricks. In August 2024 it announced a $3.5 million seed round led by Gradient Ventures. The product later shifted toward agentic investigation for fraud, dispute, AML, and KYB workflows sold to regulated financial institutions.
Compiled by commissioned research from 8 cited public sources — announcements, filings, and press listed under research sources below.
Key figures
latest reportedCompany-reported or press-reported figures, each dated to when it was claimed — not independently audited.
Competitors · 6
by search overlap
Companies competing with Roe AI for the same Google search keywords, organic and paid, via search-intersection analysis.
Timeline · 3
launches, deals, and filingsRoe published a customer story describing how Dutchie's compliance team uses Roe to automate the high-volume first pass of reviews and redirect analysts to higher-value risk work.
Gradient Ventures led Roe AI's seed funding round with participation from Ardent Ventures, Y Combinator, Orange Collective, key executives from Snowflake, and data leaders including Gu Xie of Group 1001 and Daniel Svonava of Superlinked.
$3.5M source ↗
Roe AI launched during Y Combinator's Winter 2024 batch as a data warehouse that applies multimodal AI agents to unstructured data — documents, images, websites, video — queryable via SQL combined with natural language, with connectors to S3, Snowflake, and Databricks.
Dated company events from announcements, filings, and press; legal rows summarize public dockets and regulator releases.
In the news
▸Research sources · 8
primary sources listed
- Roe AIroe-ai.com · web
8 public sources were cited for this profile; the first-party ones are listed here.
Frequently asked questions
- What does Roe AI do?
- Roe AI builds AI agents that investigate fraud, dispute, and AML cases end-to-end for financial risk and compliance teams.
- Who founded Roe AI?
- Roe AI was founded by Richard Meng.
- Who are Roe AI's investors?
- Roe AI's investors include Ardent Venture, Gradient Ventures.
- Where is Roe AI headquartered?
- Roe AI is headquartered in San Mateo, US.