Unit21
Also known as: Unit21, Inc.
Agentic AI platform for fraud and AML operations whose agents, configured to each institution's procedures, run the financial crime lifecycle from alert triage and investigation to regulator ready narratives, under human approval gates with a work log of every step and evaluation before deployment.
Unit21 is an agentic AI platform for fraud and anti money laundering operations at banks, fintechs and crypto companies, used by more than 200 customers including Chime, Intuit, Green Dot, Sallie Mae, Brex, Crypto.com and FIS.
Relaunched in March 2026 as AI Risk Infrastructure, it runs configurable AI agents inside the system where detection and investigation already live: an agent ingests an alert from real time transaction monitoring or screening, gathers evidence across the alert's own data and connected records, expands the entity network, researches external sources for adverse media and sanctions exposure, recommends a close or escalation or takes the workflow action under the customer's configured guardrails, and drafts a regulator ready narrative in the customer's format for QA and approval, with a transparent work log of every step taken, data accessed, checks performed and assumptions made.
Agents are configured to each institution's standard operating procedures, thresholds, escalation rules and review controls, and a team can build new agent tasks in plain English, describing an online search or a data analysis question, testing the task against a real alert and seeing the actual output before saving it.
Every AI decision is traceable and mapped to policy, humans make the call by default, evaluation precedes deployment and tracing follows it, customer data is never used to train the vendor's agents, and dispositions feed back into detection improvements that the platform recommends with validation.
The underlying platform covers real time monitoring, device intelligence, a fraud consortium, transaction monitoring, case management, payment and sanction screening, customer risk rating and regulatory filings, and ingests data from more than thirty named systems across core banking, digital banking, banking as a service, data warehouses, identity and crypto analytics providers, with bidirectional syncing to Salesforce and Zendesk. Unit21 holds SOC 2 Type I and Type II attestations and is sold by demo.
Vendor details
Canonical URL
https://www.unit21.ai
Category
Enterprise operations agent
Subcategory
Agentic AI for fraud and AML financial crime operations
Funding status
Independent, based in San Francisco, founded in 2018 by chief executive Trisha Kothari and co founder Clarence Chio on the premise that risk and compliance teams should not need engineering resources to build fraud and AML rules. Unit21 raised ninety two million dollars from investors including Tiger Global, Google's Gradient Ventures, and ICONIQ, and serves more than two hundred customers including Intuit, Chime, and Green Dot. In March 2026 it relaunched as AI Risk Infrastructure with a platform rebuilt around AI agents, reporting over one hundred customers already running the agents in production.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Unit21 is built on a data agnostic architecture that connects to core banking systems and payment rails, ingesting transactions, behavioral signals, device data, and customer attributes in real time from more than thirty named systems, with bidirectional syncing of cases to Salesforce and Zendesk. The unified data model brings fraud, AML, enhanced due diligence, and sanctions into one connected environment with entity networks. Developer documentation is password gated, and no public API reference is readable.
In practice
Up to ninety five percent of your AML alerts are false positives, and analysts burn days manually working them and drafting SARs. Unit21's agent ingests the alert, gathers evidence, drafts a regulator ready narrative, and logs every step, cutting investigation time dramatically.
Fraud and AML run in separate systems, so criminals exploit the blind spots between them. Unit21 unifies fraud, AML, sanctions, and case management on one data model with entity graphs that surface money mules and networks.
You need to tighten detection without months of engineering and vendor rule cycles. Unit21's AI recommends optimized rules from your alert dispositions, and you validate them with backtesting and shadow mode before deploying with confidence.
Sources & related URLs
Agentic Index coverage score
9.0 / 14 capabilities · 64%
| Integrations & Tool Calling | Full |
|---|---|
|
The integrations page publishes a named catalog of more than thirty systems across eleven categories: core banking (Fiserv DNA, Finxact and Premier, Jack Henry SilverLake and Symitar, Corelation), digital banking (Alkami, Q2, Finastra, NCR), banking as a service (Unit, Treasury Prime), data warehouses (Snowflake, BigQuery, Databricks, Aurora), KYC and KYB providers, watchlists, crypto analytics, device intelligence, check images, and Salesforce and Zendesk with bidirectional syncing of case records and investigation outcomes, plus a proprietary ingestion layer described as integrating with any tool. SourceUnit21, unit21.ai/integrationsread 2026-09-06 |
|
| Workflow Orchestration | Full |
|
Agents run the financial crime lifecycle end to end across alert triage and evidence gathering, case investigation and entity network expansion, narrative drafting and filing preparation, QA checks against SOP steps and detection improvement, investigating step by step and taking or recommending the workflow action, with a task builder through which a team defines new agent tasks and workflows configured to its own escalation rules. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
| Knowledge Grounding & RAG | Full |
|
Agents investigate over the platform's own data model, expanding entity networks during case investigation, querying the alert's own data through the Data Analysis task, and drawing on transaction, customer and case history ingested from core banking, warehouse and KYC systems into the platform; that store persists across alerts, scales past a context window and stays queryable. SourceUnit21, unit21.ai/products/ai-agent and unit21.ai/integrationsread 2026-09-06 |
|
| Human Oversight & Guardrails | Full |
|
The agents keep a human in the loop by default. Each agent run recommends a close or escalation or takes the workflow action directly according to the customer's configured guardrails, the customer sets clear guardrails for when a human must review or approve and configures approval gates, narratives are drafted ready for QA and approval, and every task builder run is approved by a human. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
| Security, Identity & Governance | Partial |
|
Unit21's security page states SOC 2 Type I and Type II attested by Armanino, with penetration tests by Doyensec and Cobalt and GDPR and CCPA compliance. The access controls the page documents are the vendor's own (Auth0 for authentication, MFA required for staff to reach production, logging of staff access to customer data). How a customer's own users are identified, provisioned or given roles is not published. SourceUnit21, unit21.ai/security and unit21.ai/products/ai-agentread 2026-09-06 |
|
| Observability & Auditability | Full |
|
Every agent run produces a transparent work log, a trace of every step taken, data accessed, checks performed and assumptions made, with the rationale for each recommendation tied to evidence and policy logic and a full audit trail of the who, what and why, and the platform traces runs after deployment. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
| Memory & State Persistence | Not documented |
|
The product page states dispositions and feedback improve future detection and decisioning over time, which is learning absorbed into the vendor's models rather than a memory layer, and the entity networks and case records the agents work over are the platform's data model. No memory with a stated scope and lifetime and no conversational session state is documented. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
| Deployment & Data Residency | Not documented |
|
The security page documents encrypted backups and network segmentation but names no hosting provider, region or customer environment option, and the documentation is password gated. Where customer data is hosted, and whether a region can be chosen, is not published. Sourceunit21.ai/securityread 2026-09-06 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
The products are modules of one platform (transaction monitoring, case management, payment and sanction screening, customer risk rating, regulatory filings), and the Agentic Task Builder lets a team build its own agent tasks in plain English. The agents are configured to each program's SOPs and thresholds rather than selected from a catalog, and the page alludes to vendor templates for agent tasks without publishing a catalog. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
| Triggers & Channel Coverage | Full |
|
The agents pick up alerts raised by the platform's real time detection and transaction monitoring, which they ingest for triage, and data arriving through the ingestion integrations from core banking, payment and warehouse systems; the product page describes an alert entering, being investigated and reaching a filing in one environment. No customer configurable schedule or webhook surface is documented. SourceUnit21, unit21.ai/products/ai-agent and unit21.ai/integrationsread 2026-09-06 |
|
| Model Flexibility & Routing | Not documented |
|
The product page states customer data is never used to train the vendor's agents and refers to its models without naming a provider, and no customer or admin model selection or vendor side routing across providers is documented on the public pages. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
| APIs, SDKs & MCP Extensibility | Partial |
|
A developer documentation portal exists at docs.unit21.ai, but every link to it resolves to a password page with no public tier. No public page names an API, SDK or MCP server through which an outside caller drives the platform; the integrations page describes the vendor's proprietary data ingestion and bidirectional syncing with Salesforce and Zendesk, which move data in and out rather than letting a caller drive the agents. SourceUnit21, unit21.ai/integrations and unit21.ai/products/ai-agentread 2026-09-06 |
|
| Testing, Debugging & Optimization | Full |
|
The product page documents evaluation before deployment and tracing after it as the answer to model risk, a task builder in which a team tests a new agent task against a real alert and sees the actual output before saving it, detection improvements recommended from outcomes with validation, and an implementation path that validates a workflow in parallel before expanding it. Backtesting and shadow mode are not described publicly. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
| Browser & Computer Use | Partial |
|
The Agentic Task Builder's Online Search task runs the external research a team describes, for example checking an entity and connected businesses for adverse media or sanctions exposure, at run time against live alerts and returns a verdict with sources. The mechanism is not stated, and no browser, desktop or computer control is documented. SourceUnit21, unit21.ai/products/ai-agentread 2026-09-06 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
Suspicious activity report autofill in Unit21 now covers more fields that teams configure themselves, including FinCEN advisory codes, values drawn from entity and instrument data, and checkboxes set from alert or case tags. Unit21 calls it the first wave of autofill coverage.
Bears on: Workflow orchestration
View sourceUnit21 introduced SAR Agents, which automatically draft FinCEN-ready narratives directly into filings for analyst review. The update also includes expanded payments screening with PEP and Adverse Media watchlists, goAML integration into the filing form generator, and enhanced auto-assignment routing.
Bears on: Agent capability
View sourceUnit21 released a new API suite that allows financial institutions to call its financial crime AI agents directly from their existing case management systems. Instead of migrating to Unit21's full platform, customers can assign investigation tasks to the agent via API and receive the decision, gathered evidence, and a written narrative as structured data.
Bears on: MCP / tool calling / API
View sourcePricing
Contact for pricing
What is public
No list pricing. Unit21 sells to banks, fintechs, and crypto companies through enterprise engagement, but rates are not public.
Billing mechanics
The billing basis is not disclosed; Unit21 quotes rates through enterprise engagement after a demo.
Cost watchouts
Cost scales with transaction and alert volume and the number of modules deployed across fraud, AML, sanctions, and case management, and adding the AI agents on top of monitoring is a distinct commercial line.
Variable cost rationale
Priced against transaction and alert volume and the modules deployed, so cost grows with monitored volume and how much of the fraud and AML lifecycle runs on the platform.
Additional watchouts
Confirm whether pricing is per transaction, per module, or a platform subscription, and how the agentic AI capabilities price relative to the base monitoring and case management platform.
Sales call required
Yes, required for paid access
Free / trial
Demo on request; no public free tier
Key ambiguities
No public rate is disclosed, and whether pricing is per transaction, per module, or a platform subscription with users is not clear, nor how the AI agents price relative to the base platform.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Adopt AI — AI CPA firm whose agents run reconciliations, close, AP and AR, and…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
- Alloy.ai — Commerce intelligence system for consumer brands that unifies four…
Alternatives to Unit21
The closest documented capability profiles to Unit21 among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Bretton AI10.0 / 14Adds documented Deployment & Data Residency
- ketteQ9.0 / 14Fuller documented coverage on Prebuilt Agents, Templates & Packs
- Nova Recruiter8.5 / 14Fuller documented coverage on APIs, SDKs & MCP Extensibility
- Optimal Dynamics7.5 / 14A lighter documented profile than Unit21
- Phaidra7.5 / 14A lighter documented profile than Unit21
- Enboarder9.0 / 14Adds documented Deployment & Data Residency
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded