Fenergo

The system of record, its regulatory rules and its case handling run without models, so the AI is the engine of particular capabilities rather than the whole product. Since July 2026 those capabilities are six KYRA automation agents: document classification and extraction with large language models, screening hit resolution, data sourcing, task completion, event significance and reporting.

Transaction monitoring adds a hybrid engine that pairs machine learning with scenario rules, and identity verification adds face matching and liveness. Fenergo says intelligence, autonomy and compound classes of agent will follow in the coming quarters.

Each KYRA agent publishes what it does alone and where it stops. KYRA:Complete closes a workflow task only when required fields, approved documents, data integrity checks and configured rules all pass, and routes anything missing or needing judgment to an analyst with context. KYRA:Docs holds conflicts for review and commits no document data until an analyst approves it.

KYRA:Screen runs in manual, semi autonomous or fully autonomous mode set by policy, risk and oversight needs, with a rationale recorded for every match decision, and the significance agent sorts each data change as significant, insignificant or needing review against set thresholds.

Every agent action, source, decision and rationale is written into Fen-X as the work happens, and KYRA:Docs links each extracted value back to its page and line in the source document, so a validator can trace what each agent did and why. Fenergo publishes no accuracy figure, extraction error rate or validation report for any agent. Dashboards show agent performance by queue, mode and jurisdiction, and new transaction monitoring rules can be back tested before release.

Aggregate figures come from Fenergo itself: 88 percent faster onboarding than the industry average and 72 percent more automation in document processing. Its agent case studies describe an Australian bank, a global bank and a North American bank without naming them. Named customers carry deployment facts instead.

TMF Group rolled the platform out to 2,000 users in more than 125 offices across 87 jurisdictions in four months, U.S. Bank Investment Services went live on Fen-X for investor onboarding with its Treasury Management business already on it, and BNP Paribas's One KYC program won a Celent Model Bank award.

Strict data isolation and no training on client data are components of the AI governance framework, and the privacy notice adds that personal information processed with AI does not train models. Every AI capability is opt in or configurable by the client, so each institution chooses what runs on its data. Amazon Bedrock sits among the AWS services Fenergo uses, which makes the no training commitment one that has to hold at the model provider too.

Fenergo acts as processor for client data, with the client as controller, names a data protection officer in Dublin and runs a privacy request portal. Its trust center lists every subprocessor publicly, nine in all, from AWS for hosting to Elastic for transaction monitoring search and Okta for portal sign in, each with its purpose, data types and region. Retention is stated only as no longer than necessary, though certificates of destruction sit among the trust center's evidence.

Fenergo's trust center, open to any visitor, lists its ISO/IEC 27001:2022 certificate, a 2025 SOC 2 Type 2 audit report, penetration test reports from February and May 2026 with a June retest, its statement of applicability and more than twenty policies, with the SOC 2 report released under a nondisclosure agreement. Schellman became its ISO 27001 certification body on 29 June 2026, replacing Amtivo. Data at rest is encrypted, typically with 256 bit Advanced Encryption Standard, and Fenergo published a statement on the Drift data breach.

Regulatory policy logic in Fen-X is maintained by Fenergo across jurisdictions, but no license, supervisor or register entry of its own is stated. Supervisors examine the banks using the platform, which is supervision of the customers rather than the vendor, and First Abu Dhabi Bank's move to the cloud platform involved working with local regulators.

Fenergo runs its AI under a SaaS AI Control Framework of more than 30 controls, applied across the AI lifecycle from design and deployment to monitoring and retirement, with bias and fairness mitigation and human oversight among its areas. An AI governance policy and AI management system documents sit in its trust center behind an access request.

The identity verification product now matches faces with active and passive liveness, and no demographic performance result for that matching is published, from Fenergo's own testing or from an independent evaluation. In screening, KYRA:Screen records a rationale for every match decision and reports performance by queue, mode and jurisdiction, so a compliance team can see where its decisions cluster.

Recorded rationales let an institution reconstruct why a client was refused, but Fenergo publishes no guarantee, indemnity or accuracy commitment for agent output. A refused person is typically not told why, a legal requirement in money laundering cases, which leaves a missed exception in a straight through case to the institution.

The subprocessor list is public in the trust center: AWS for hosting, Elastic, Okta, Sinch, Atlassian, Gainsight, Trustwave, Microsoft for Sentinel and Intercom, each with its purpose and region. Which foundation model the KYRA agents run on is unstated, and so are the registries and screening lists the data sourcing agent compares. Amazon Bedrock is named among the AWS services Fenergo uses, which places the model layer inside AWS.

Fen-X is the legal entity system of record, so onboarding and review run on data held in Fenergo's own platform. Integration Hub manages the standard integrations from one dashboard, and Integration Flows lets clients build their own on Fenergo's infrastructure with a visual editor. KYRA reaches Fen-X data through the Model Context Protocol and takes requests from approved outside agents through its Agent to Agent interface. Named partners include LexisNexis Risk Solutions, Moody's, LSEG and Salesforce, and a Fenergo app for Salesforce is sold on AWS Marketplace.

Customers choose the region for their tenant: the US, Canada, the EU, Singapore, the United Arab Emirates or Australia, on multi tenant software as a service run on AWS across three availability zones per region. Subprocessors such as Elastic and Okta follow the selected region. First Abu Dhabi Bank configured external key management and encryption on the platform to meet its regulators' requirements.

Fenergo's AWS Marketplace listing prices 12 month contracts at $150,000 for a Small deployment, $350,000 for Medium and $850,000 for Enterprise, all through private offer. Whether the Fen-AI agents sit inside those tiers is not stated. The tier follows how many jurisdictions and lines of business the deployment covers, pricing can vary with the types and volumes of client entities, and partner implementation and AWS infrastructure costs come on top.

Segment offers are maintained for corporate and institutional banking, commercial and business banking, retail banking, private banking, asset management, asset servicing, fintechs and energy and commodity trading. Named institutions span them: LBBW in corporate banking, Ninety One and Union Investment in asset management, U.S. Bank Investment Services and Northern Trust in fund services, and First Abu Dhabi Bank in the Gulf. Workflow coverage runs from onboarding through transaction monitoring and periodic review to offboarding.