FIS offers its AML agent, built with Anthropic, to any US bank on a design partner's projections
FIS has made its AML investigations agent, built with Anthropic, available to any US bank, with Amalgamated Bank deploying it. The performance figures so far are the bank's projections, including a 98% cut in Level 2 case time.
The Investor · Invest desk
Bar comparison of Amalgamated Bank's projected time reductions: Level 2 case investigation about 98%, Level 1 alert review more than 90%, and suspicious activity report narrative preparation more than 80%.
Amalgamated Bank projections, % reduction in time per task In % less time
| Item | Value | Claim |
|---|---|---|
| Level 2 investigation, approx. | 98% less time | 7 |
| Level 1 alert review, over | 90% less time | 6 |
| Report narrative prep, over | 80% less time | 6 |
What happened
- Amalgamated Bank was a design partner during development, and the agent is the first product in FIS's planned roadmap of AI agents for banking.
- Amalgamated projects Level 1 alert review time falling by more than 90% and suspicious activity report narrative preparation time by more than 80%.
- Before assembling evidence, the agent maps each alert against the customer's type, historical activity and expected behaviour, so the summary reflects what is normal for that customer.
- FIS developed and owns the agent, which runs in FIS's private cloud on Anthropic's Claude models, with an Anthropic team embedded in FIS's product and compliance teams.
- The agent has native integrations with FIS's Horizon and Systematics core banking platforms; FIS IBS comes after them on the roadmap.
Why it matters
- What it limits constraint Every case still ends with an analyst's confirmation, so any saving has to come from shorter reviews of each case.
- Who's exposed exposure Adopting banks will have the agent's recommendation, its reasoning and the analyst's final decision in one audit trail that internal audit and regulators can examine.
- Who has to decide decision A bank outside FIS's client base has to decide whether to buy on a design partner's projections or wait for a measured Level 2 figure.
- What it sets up precedent FIS says it will reuse the evaluation frameworks and engineering practices from this build for further agents, so Amalgamated's results will set expectations for each one that follows.
Three outcomes are plausible once Amalgamated Bank has live numbers: they match the projections, they fall well short, or banks on other core systems find the integration list too short to bother. We'd put money on the second, and the reason starts with the 98%.
AnChain.AI's 2025 research puts evidence assembly at 40% to 60% of investigation time [4]. An agent that only assembled evidence, and did it instantly, would cut 60% at most. Amalgamated projects about 98% for Level 2 case investigation [7], which is 38 percentage points above that ceiling [20], and on that projection 100 hours of Level 2 work becomes 2 [21]. The comparison is loose, since it sets an industry average against one bank's Level 2 queue, and the bank's cases may be unusually heavy on assembly. Or the agent shortens the analysis as well. We'd still expect the measured figure to come in lower, because the bank built its projections by comparing current workflows with the agent's expected capabilities, and they are not results from live deployment [8]. A measured Level 2 saving above 90% would prove us wrong.
The Level 1 claim has a different shape. Datos Insights put industry false-positive rates at 90% to 95% in 2023 [5], which leaves 5 to 10 genuine alerts in every 100 [22]. A Level 1 cut above 90% [6] therefore depends on clearing the false ones quickly without burying the genuine few, and the customer-context step is the feature aimed at that [9].
FIS is not selling a replacement for the systems that generate alerts. The agent works alongside banks' existing transaction monitoring and case management [12], and each institution's risk logic, thresholds and procedures stay isolated in the FIS Control Plane [13]. Stan Szwalbenest, FIS's head of enterprise fraud and financial crimes, said what holds investigations up is evidence spread across many systems and alerts that all get the same treatment; analyst judgement is not what is missing [17]. Sean Searby, Amalgamated's chief information and operations officer, noted that investigators remain accountable for final determinations [18].
The Paypers' report does not state pricing, name any bank other than Amalgamated as a user, or describe how a bank on a core system other than Horizon or Systematics would connect [19].
What to watch
- Measured results from Amalgamated's live deployment against the projected cuts of more than 90% at Level 1, about 98% at Level 2 and more than 80% in SAR narrative preparation.
- Whether a US bank with no FIS core system signs up, and whether FIS IBS integration arrives as the roadmap says.
- Which task FIS picks for its second agent, and whether it reports measured results from this one first.
Information, not investment advice. This story is general information about companies and markets, drawn from cited reporting. It is not a recommendation to buy, sell or hold any security; check the cited sources before acting.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence30
- Adoption12
- Hype gap+40
- Incentives65
- Confidence45
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
As of October 2026, the FIS AML Investigations Agent is available to any US-based bank, including institutions that are not existing FIS clients.
- [2]
The agent is currently being deployed at Amalgamated Bank.
- [3]
Amalgamated Bank acted as a design partner during development, and the agent is the first product in FIS's planned roadmap of AI agents for banking.
- [4]
Research from AnChain.AI (2025) indicates that analysts spend 40% to 60% of investigation time assembling data before analysis begins.
- [5]
Datos Insights (2023) puts industry false-positive rates at 90% to 95%.
- [6]
Amalgamated Bank projects that the agent will reduce Level 1 alert review time by more than 90% and suspicious activity report narrative preparation time by more than 80%.
- [7]
Amalgamated Bank projects a reduction of approximately 98% in Level 2 case investigation time.
- [8]
Amalgamated Bank compared its current investigation workflows against the agent's expected capabilities to arrive at its projections; the figures are projections rather than measured results from live deployment.
- [9]
Before assembling evidence, the agent maps each alert against the customer's type, historical activity and expected behaviour, so that the resulting summary reflects what is normal for that customer.
- [10]
All of the agent's outputs are recommendations, and no case can be closed without explicit analyst confirmation.
- [11]
Recommendations, the reasoning behind them, and the analyst's final decision are logged in an audit trail intended to support internal audit and regulatory examination.
- [12]
The agent is designed to work alongside banks' existing transaction monitoring and case management systems rather than replace them.
- [13]
Each institution's risk logic, thresholds and procedures are kept isolated through the FIS Control Plane, FIS's governance layer for agent access and institution-specific configuration.
- [14]
FIS developed and owns the agent, which runs in FIS's private cloud and uses Anthropic's Claude models; an Anthropic team worked embedded within FIS's product and compliance teams.
- [15]
According to FIS, the collaboration with Anthropic also produced evaluation frameworks and engineering practices that FIS will use to build and scale further agents on the same governed platform.
- [16]
The agent integrates natively with the FIS Horizon and FIS Systematics core banking platforms, and FIS IBS is next on the roadmap.
- [17]
Stan Szwalbenest, Head of Enterprise Fraud and Financial Crimes at FIS, said that investigations are slowed by evidence scattered across systems and by uniform treatment of alerts, not by a lack of analyst judgement.
- [18]
Sean Searby, Chief Information & Operations Officer at Amalgamated Bank, noted that investigators remain accountable for final determinations.
- [19]
The Paypers' report on the announcement does not state pricing, name any bank other than Amalgamated Bank as a user, or describe how a bank on a core system other than Horizon or Systematics would connect.
- [20]
The projected 98% Level 2 reduction sits 38 percentage points above the 60% upper bound of time spent assembling evidence.
- [21]
On the projected 98% cut, 100 hours of Level 2 investigation work would fall to 2 hours.
- [22]
With false-positive rates of 90% to 95%, between 5 and 10 of every 100 alerts are genuine.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
3 independent publishers whose own reporting we read for this story.
- fintech.globalFIS opens AML agent to non-clients as alert volumes soar
1 article · October 16, 2026
- pymnts.comFIS and Anthropic Debut Money Laundering Investigation Agent
1 article · October 15, 2026
- thepaypers.comFIS opens AML investigations agent built with Anthropic to US banks
1 article · October 16, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Financial crime complianceFollow
- Agentic AI in Banking OperationsFollow
- Anti-Money LaunderingFollow