Vendor comparison
AI compliance agents compared: Castellum.AI, Bretton, Sphinx, WorkFusion and Zyphe
Five vendors that sell AI agents for AML operations, side by side. Four of them frame their compliance mainly around United States bank supervision. If you are supervised in the EU, that is the comparison that decides it.
In short: this page compares five vendors that sell AI agents for AML operations to financial institutions: Castellum.AI, Bretton (formerly Greenlite AI), Sphinx, WorkFusion and Zyphe. All five use agents to review alerts and prepare or resolve cases, and all five say they document the reasoning behind each decision. The difference is which supervisor they were built for. Castellum.AI, Bretton, Sphinx and WorkFusion frame their regulatory posture mainly around United States supervision, citing the OCC, the FDIC, the Federal Reserve, the NCUA, FinCEN, OFAC, the New York Department of Financial Services and the California DFPI on their own websites; WorkFusion also cites the Wolfsberg Group's principles for AI. None of the four publicly addresses Article 18 of Regulation (EU) 2024/1624, the EU Anti-Money Laundering Regulation, which governs how an EU obliged entity may outsource AML work from 10 July 2027. Zyphe is built around that Article: every desk states which tasks it will not accept, supports supervisor pre-notification, and ships a written rationale with each case.
Vendor facts checked on 1 October 2026. Written by Michelangelo Frigo, Co-Founder and CEO at Zyphe.
How we compared
Seven criteria, applied the same way to every vendor, and each competitor fact linked to the vendor page it came from.
Zyphe wrote this comparison and is one of the vendors in it. Facts about other vendors come from their own public pages or the sources cited, as of this page's last update on . No vendor pays to appear here (editorial policy). Out of date? Tell us. Our own entry, limitations included: Where Zyphe fits.
- 1. Supervisory framing: which supervisors and rules the vendor names on its own site.
- 2. AMLR Article 18: whether the vendor addresses outsourcing under Regulation (EU) 2024/1624 and publishes the tasks it will not take on.
- 3. Decision rationale: whether each decision comes with written reasoning a reviewer or supervisor can read.
- 4. Human decision boundary: who makes the call by default, as the vendor describes it.
- 5. Data protection and deployment: published certifications and commitments, and where the vendor says data sits.
- 6. Integration and delivery: how the agents connect to your systems, and whether the vendor sells an operated or managed service.
- 7. Data and evidence: whether the vendor brings its own watchlist data, and which customers it names.
Vendor facts checked on , against each vendor's own website, documentation and press pages; nothing comes from a demo, a private conversation or a customer. Zyphe wrote this page and is one of the five vendors in it, so our own column links the Zyphe page each statement rests on. How we source and review comparisons is in our editorial policy. If a cell is out of date, tell us and we will correct it in public.
What does each vendor actually sell?
Positioning quoted from each vendor's own homepage, so nothing here is contested.
Castellum.AI
"Agentic Screening and Alert Resolution"
Screening and automated alert resolution built on watchlist data Castellum.AI compiles itself, with a family of agents branded Arbiter for sanctions and PEP alerts, adverse media, transaction monitoring and fraud, and an AI Governance Center that sets out how the agents are tested, validated and monitored. Of the five, it has the strongest claim to owning the underlying data rather than renting it.
Bretton
"AI-native operations for the financial back office"
Rebranded from Greenlite AI in February 2026, and greenlite.ai now redirects to Bretton. Scoped wider than AML: a template library, a natural-language agent builder and a shared workbench for back-office work generally, plus AI-native managed services for AML that pair the platform with named onshore investigators. It publishes named case studies with banks and fintechs and offers virtual private cloud deployment.
Sphinx
"Your last compliance hire"
Agents that sign in to the systems you already run and work them the way an analyst would, so no API integration or engineering time is required, alongside Frontline, an AI-native BPO whose embedded analysts work your queue with the agents, and a document fraud detection product.
WorkFusion
"AI Agents for financial crime compliance"
Named, pre-built agents for Level 1 alert review and Level 2 investigation work: Evelyn for name screening, Tara for transaction screening, Evan for adverse media, Edward for enhanced due diligence, Isaac for transaction monitoring and Kayla for KYC. They clear false positives, escalate the rest to analysts and connect to screening engines such as World-Check, Firco and Bridger. WorkFusion is a UiPath company, and since 30 September 2026 its agents are also offered as a managed service through UiPath partners including Capgemini, Deloitte, Mphasis and Protiviti. It names Scotiabank, Deutsche Bank and Standard Bank among its customers and says it serves 10 of the top 20 banks.
Zyphe
Operated compliance desks, agent-prepared and human-approved
Desks that work a named queue: sanctions and PEP alerts, adverse media, UBO and EDD, transaction monitoring triage, periodic KYC refresh. By default agents prepare the file and your team approves the decision; Auto-pilot runs only on categories your organisation has validated and switched on, an agent's action goes through only if a rule-based check also passes, and your MLRO keeps final sign-off. Every case ships with a written rationale, and there is no central store of customer PII.
How do the five compare?
Every competitor cell reflects what the vendor publishes about itself, not what we infer about it, and links the page it came from.
| Criteria | Castellum.AI | Bretton | Sphinx | WorkFusion | Zyphe |
|---|---|---|---|---|---|
| Supervisory framing on its own site | OCC, NCUA, NY DFS, California DFPI (US) Homepage | OCC, FDIC, Federal Reserve (US) Homepage | FDIC, OCC, Federal Reserve (US); its blog also discusses the EU AML Regulation Homepage · Blog post on KYB platforms | BSA and OFAC; cites the Fed, FDIC, FinCEN, NCUA and OCC (US) and the Wolfsberg Group principles for AI Sanctions screening page | AMLR, AMLA, EU national supervisors How the desks work |
| Addresses AMLR Article 18 outsourcing | Not stated publicly | Not stated publicly | Not stated publicly | Not stated publicly | Yes, on every desk How the desks work |
| Publishes which tasks it will not accept | Not stated publicly | Not stated publicly | Not stated publicly | Not stated publicly | Yes, against Article 18(3) How the desks work |
| Written rationale with each decision | Yes: rationale, data references and a human-readable narrative for every agent decision Arbiter page | Yes: every decision logged with reasoning, evidence and model version Homepage | Yes: step-by-step reasoning, backed by evidence Homepage | Yes: each decision documented, with written explanations and audit trails Homepage · Tara agent page | Yes: sources reviewed, reasoning applied and action taken, logged per decision Security |
| Who makes the call by default | Your policies set which outcomes agents accept and which go to human review AI Governance Center | Agents draft the full case for your team to approve Platform page | Agents resolve and route cases on your risk logic Homepage | Agents clear false positives and escalate the rest to your analysts Tara agent page · Evelyn agent page | Your team approves by default. Auto-pilot runs only on categories you have validated and switched on, an agent's action goes through only if a rule-based check also passes, and your MLRO keeps final sign-off Security |
| Published security and data-protection posture | SOC 2 Type II Arbiter page | SOC 2 Type II, GDPR, zero-data retention, independent model validation Homepage · Trust center | SOC 2 Type II, GDPR Homepage | SOC 2 Type II; a data processing agreement with the EU Standard Contractual Clauses and the UK Addendum Kayla agent page · Data processing agreement | No central store of customer PII. SOC 2 and ISO/IEC 27001 audits are in progress, with completion targeted Q4 2026. Security · GDPR |
| Deployment and where data sits | Not stated publicly | Virtual private cloud deployment offered Homepage | Not stated publicly | Cloud service, or the Work.AI platform installed on your own servers Data processing agreement · Installation docs | Documents encrypted and split across thousands of storage nodes into the end user's own vault; EEA and UK end-user data stays on EEA and UK nodes GDPR |
| Integration model | API and data feeds; agents added as a layer in your case management, core or screening system Arbiter page | Platform overlay with templates, builder and workbench, and native connectors to your systems Platform page · Homepage | Agents sign in to your existing systems; no API needed Homepage | Pre-built agents on your screening and case systems, with ready integrations for World-Check, Firco and Bridger Evelyn agent page · Tara agent page | Operated desks inside your existing tools How the desks work |
| Operated or managed service | Not stated publicly | Yes: AI-native managed services for AML, with named onshore investigators on judgment calls Managed services page | Yes: Frontline, embedded analysts working your queue with the agents Frontline page | Through UiPath partners, including Capgemini, Deloitte, Mphasis and Protiviti Managed services announcement, 30 September 2026 | Yes: operated review desks priced per completed task, with no monthly minimum on pay as you go Desk pricing |
| Own watchlist data | Yes, compiled in house from 200,000+ sources Homepage | 180+ connected data sources Homepage | Not stated publicly; runs its own real-time risk network for counterparty links Homepage | No: reviews the alerts your existing screening system raises Evelyn agent page | No: works the alerts your existing screening engine produces Sanctions and PEP desk |
| Named customer references on its site | A Lead Bank case study and testimonial; an anonymised Fortune 50 case study Guides and case studies | Named case studies, including Payward, First Internet Bank and Grasshopper Bank Case studies | Named case studies, including Conduit, Equals and Alviere Case studies | Named customer stories and quotes, including Scotiabank, Deutsche Bank, Standard Bank and Valley Bank Homepage · Customer stories | Identity verification customer stories; agent desk deployments referenced privately, on request Customer stories |
What "not stated publicly" means
We did not find it on the vendor's public website, documentation or press pages when we checked on 1 October 2026. It is not a claim that the vendor cannot do the thing, and it is not a claim that it does the thing badly. All four are credible operators, and against a United States supervisor all four are better referenced than we are.
On zero retention
Bretton publishes a zero-data-retention commitment, and WorkFusion's data processing agreement commits it to deleting or returning customer data when the contract ends, so we are not claiming to be the only vendor that limits what it holds. A retention policy is a promise you verify through contract and audit. Our position is about architecture: documents are encrypted and split across storage nodes into the end user's own vault, so there is no central store of customer PII at Zyphe to breach. We do keep verification results, audit logs and cryptographic proofs (security, GDPR).
Sources, each checked on 1 October 2026:
- Castellum.AI: Homepage, Arbiter page, AI Governance Center, Guides and case studies.
- Bretton: Homepage, Platform page, Trust center, Managed services page, Case studies, Newsroom. The newsroom lists the Greenlite AI rebrand, and greenlite.ai redirects to Bretton.
- Sphinx: Homepage, Blog post on KYB platforms, Frontline page, Case studies.
- WorkFusion: Sanctions screening page, Homepage, Tara agent page, Evelyn agent page, Kayla agent page, Data processing agreement, Installation docs, Managed services announcement, 30 September 2026, Customer stories, AI Agents page, Company page.
- Regulation (EU) 2024/1624: quoted from EUR-Lex.
What does AMLR Article 18 require when you outsource AML work?
This is the whole argument for choosing an EU-built vendor, so it is worth stating from the primary text rather than from a summary.
Article 18 of Regulation (EU) 2024/1624 is titled Outsourcing. It permits outsourcing and attaches conditions. Article 18(1) requires you to notify your supervisor before the provider starts work. Article 18(2) treats the provider as part of your organisation, leaves you fully liable for the outcome, and requires you to demonstrate to your supervisor that you understand the rationale behind the outsourced activity. That last requirement is the reason a Zyphe case ships with written reasoning rather than a score: a confidence number is not something an MLRO can explain to a supervisor. The Article applies from 10 July 2027.
Which tasks can never be outsourced?
- The proposal and approval of your business-wide risk assessment
- The approval of your internal policies, procedures and controls
- The decision on the risk profile attributed to a customer
- The decision to enter into a business relationship or carry out an occasional transaction
- Reporting suspicious activity, or threshold-based reports, to your FIU
- The approval of the criteria used to detect suspicious or unusual transactions
Two qualifications belong with that list. Article 18(3)(e) permits FIU reporting to be outsourced to another obliged entity in the same group and established in the same Member State. Article 18(7) lets a collective investment undertaking without legal personality outsource points (c), (d) and (e) to one of its service providers, once it has notified its supervisor and the supervisor has approved. Neither reaches a typical bank, payment institution, electronic money institution or crypto-asset service provider.
Approving an individual alert disposition is not itself an Article 18(3) requirement. The Regulation governs the decisions listed above; how much of the rest your team reviews is set by your own policy. By default Zyphe returns every case for approval, because that is the service boundary we sell, not because a regulation compels it case by case. Auto-pilot runs only on categories your organisation has validated and switched on, an agent's action goes through only if a rule-based check also passes, and your MLRO keeps final sign-off.
When should you choose one of the others?
A comparison page with no losing rows is an advertisement. These are the cases where we are not the right answer.
Castellum.AI
Choose Castellum.AI if your real problem is screening data quality and you want the list and the agent from one vendor.
Bretton
Choose Bretton if you are a large institution automating beyond AML, you need a virtual private cloud deployment, or you want a managed service with onshore investigators to replace an AML BPO.
Sphinx
Choose Sphinx if you need throughput next week and cannot free up engineering for an integration, or you want embedded analysts working your queue alongside the agents.
WorkFusion
Choose WorkFusion if you are a large bank with an established screening engine and want pre-built agents with a record at large banks, or you want a large integrator such as Capgemini or Deloitte to run the operation.
Zyphe
Choose Zyphe if you are an EU obliged entity and your supervisor will ask who did the work and on what basis, or you want customer documents kept out of a central vendor database.
Where Zyphe fits
Zyphe's own entry, identical on every comparison we publish.
- Best for
- Regulated teams that keep their KYC, KYB and AML tools but want an AI investigation layer that prepares decision-ready, evidence-backed case files.
- How it works
- Agents prepare KYC, KYB, AML and sanctions reviews with the evidence and a written rationale attached. By default your team approves each one, and your MLRO keeps final sign-off.
- Decision boundary
- The decisions AMLR Article 18(3) reserves for the obliged entity, such as the customer risk profile and whether to onboard, stay with your team. How the review desks work
- How to buy
- Agent review desks, priced per completed task with no monthly minimum, or a Growth or Enterprise plan; or the verification platform and API, priced per verification. Desk pricing · Platform pricing
- Integration
- A no-code link, a hosted flow, or the REST API with signed webhooks; SDKs for Node.js, browser JavaScript and React Native. Integration docs
- Limitations
- SOC 2 and ISO/IEC 27001 audits are in progress, with completion targeted Q4 2026. KYB register coverage depends on what each government exposes, and director data is often missing from official sources.
- Evidence
- Customer stories · Sample KYB report
What has Zyphe actually run?
Figures from live deployments. References are available under NDA.
- screening alerts discounted per agent per day
- 600 screening alerts discounted per agent per day
- reviews handled across deployments
- 120,000+ reviews handled across deployments
- operational cost saved across deployments
- $3.3M operational cost saved across deployments
Published evidence: customer stories (identity verification deployments; desk deployments are referenced privately), a sample KYB report showing what a finished file contains, and how data is handled on the security and GDPR pages.
Frequently asked questions
Who are the main alternatives to Castellum.AI, Bretton, Sphinx and WorkFusion?
Vendors selling AI agents for AML operations as of October 2026 include Castellum.AI, Bretton (formerly Greenlite AI), Sphinx, WorkFusion and Zyphe, alongside automation vendors such as Silent Eight and screening providers such as ComplyAdvantage. They differ less on what the agents do than on which supervisor their public material is written for, where they say the human decision sits, how they are delivered, and whether the vendor states where its own responsibility stops.
Which AI compliance agent vendors are built for EU AML rules?
On their own public websites as of 1 October 2026, Castellum.AI, Bretton, Sphinx and WorkFusion frame their regulatory posture mainly around United States supervision, and none of them addresses outsourcing under Article 18 of Regulation (EU) 2024/1624. Sphinx's blog discusses the Regulation's KYB requirements, and WorkFusion says four top-10 EU banks use its automation. Zyphe is built around the Regulation: each desk states the tasks it will not accept under Article 18(3), supports the supervisor pre-notification that Article 18(1) requires, and returns a written rationale per case.
What does AMLR Article 18 require when you outsource AML work?
Article 18 of Regulation (EU) 2024/1624 permits outsourcing and attaches conditions. Article 18(1) requires you to notify your supervisor before the provider starts. Article 18(2) treats the provider as part of your organisation, leaves you fully liable, and requires you to demonstrate to your supervisor that you understand the rationale behind the outsourced activity. Article 18(3) lists tasks that cannot be outsourced. The Article applies from 10 July 2027.
Which AML tasks can never be outsourced under AMLR?
Article 18(3) names six: proposing or approving your business-wide risk assessment, approving internal policies and controls, deciding a customer risk profile, deciding whether to enter a business relationship, reporting to your FIU, and approving the criteria used to detect suspicious transactions. Two qualifications belong with that list, in Article 18(3)(e) and Article 18(7), and neither reaches a typical bank, payment institution, EMI or crypto-asset service provider.
Is an AI agent reviewing my alerts an automated decision under GDPR Article 22?
It depends on who decides. GDPR Article 22 restricts decisions based solely on automated processing that produce legal or similarly significant effects, and the CJEU reading in SCHUFA is broad. By default Zyphe agents prepare the file and a named person on your team approves the outcome. Your organisation can switch on Auto-pilot for a category of case only after validating the agents on it; an agent's action then goes through only if a rule-based check also passes, and your MLRO keeps final sign-off. Whether to switch Auto-pilot on is part of your assessment as controller, and it is what your DPO will ask about first.
What is the difference between a zero-retention policy and no central store of customer PII?
A retention policy is a commitment about what a vendor chooses to keep, and you verify it through contracts, audits and trust. No central store is a statement about architecture: Zyphe encrypts documents and biometrics and splits them across storage nodes into the end user's own vault, so they are not held in one Zyphe database, while Zyphe keeps verification results, audit logs and cryptographic proofs. Bretton publishes a zero-data-retention commitment, and WorkFusion's data processing agreement commits it to deleting or returning customer data when the contract ends, so Zyphe is not the only vendor limiting what it holds. The difference is the kind of promise, not the direction of it.
Related
- The operated desks
What each desk works, what it hands back, and where the Article 18 boundary sits.
- Hiring versus agents versus BPO
The other comparison: cost, delay and coverage of another internal hire against a managed operating model.
- Switching KYC providers
Migration steps, contract timing, and what changes when the new provider keeps no central store of PII.
Book a demo
Bring the shortlist. We will take the hardest queue on it.
Book a demo with your real alert backlog. We will run a slice of it, hand back the cases with written rationales, and you can compare the output against anyone else on your list.