The company teaching AI to explain itself - clearing millions of compliance alerts, and writing down exactly why for each one.
Every large bank runs a machine most customers never see. It scans names against sanctions lists, watches transactions for laundering, and flags anything that looks off. The problem is scale: the machine throws off hundreds of thousands of alerts, and the overwhelming majority are false alarms. Silent Eight builds the artificial intelligence that reads those alerts, decides which ones are real, and - crucially - writes down the reasoning a human and a regulator can check.
Founded in 2013 by Martin Markiewicz, Julia Markiewicz and Michael Wilkowski, the company set out with a narrow mission: apply purpose-built AI to financial crime compliance. Its flagship product, the Iris platform, now spans the full compliance lifecycle - screening, investigation, transaction monitoring, case documentation and quality assurance. The latest version, Iris 7, introduces policy-bound AI agents that work within a bank's own rules rather than around them.
Silent Eight builds purpose-built AI for financial crime compliance - trained not on a single large language model, but on smaller, task-specific models.
Compliance teams spend most of their day on triage. An analyst opens an alert, checks it against sanctions and adverse-media data, decides whether it is a genuine risk, and documents that decision for auditors. Repeat, thousands of times. The bottleneck is not detection - it is resolution. Silent Eight aims squarely at that bottleneck.
Screening systems flag far more than can be reviewed. Iris lowers alert volumes without lowering risk coverage.
Regulators cannot accept a decision they cannot read. Silent Eight's neuro-symbolic models explain, in words, why an alert was raised or closed.
Every decision needs a documented rationale. Iris drafts auditable case narratives automatically, turning decisions into records.
The obvious move in 2023 was to wire everything to one large language model. Silent Eight did not. Instead it trains a collection of small, task-specific models - one whose only job, for example, is knowing that a single person's name can be spelled a dozen different ways as it travels across languages and alphabets.
Paired with symbolic reasoning, that approach lets the AI both detect risk and justify its conclusion. In a field where an unexplained decision is a liability, transparency is not a feature bolted on at the end - it is the product. That is also why two of the biggest banks in the world were willing to embed it.
Unifies screening, investigation, monitoring, documentation and QA across the compliance lifecycle, with policy-bound AI agents.
Neuro-symbolic AI that checks customer and counterparty names against sanctions and watchlists, resolving and explaining each alert.
Automated adjudication of payment screening alerts - cutting false positives while holding risk coverage.
Behavioral analysis that surfaces anomalies and customer-risk patterns earlier in the transaction flow.
Generates auditable case narratives and investigation write-ups, turning a decision into a regulator-ready record.
Automated sampling, model validation and screening audits for full oversight - no black boxes.
Silent Eight sells to compliance, sanctions and financial-crime operations teams inside large financial institutions. Standard Chartered deployed its AI in sanctions operations as far back as 2018; HSBC signed a multi-year partnership in 2021 and folded the technology into its global compliance infrastructure. Deployments have since expanded across Gulf banks including Emirates NBD, Mashreq, First Abu Dhabi Bank and Abu Dhabi Islamic Bank.
Two of Silent Eight's biggest customers, HSBC and Standard Chartered, are also investors. When the people using your product write you a check, you have solved a real problem.
Business model: B2B enterprise SaaS. Silent Eight licenses the Iris platform and its AI adjudication models to banks under multi-year contracts, integrating with the systems they already run - a quiet infrastructure play rather than a consumer brand.
| Round | Amount | Date | Notable investors |
|---|---|---|---|
| Seed | Undisclosed | Sep 2015 | Wavemaker Partners |
| Series A | $6.2M | Nov 2019 | OTB Ventures, Wavemaker Partners, SC Ventures |
| Series B | $40M | Mar 2022 | TYH Ventures (lead), HSBC Ventures, SC Ventures, Aglaia Family Office |
The 2022 Series B brought total capital raised to roughly $55M. Valuation was not disclosed but reported to quadruple the October 2020 figure.
Mathematician and serial founder across Europe and Asia; leads strategy and the company's bet that purpose-built AI can tackle financial crime.
Runs operations, scaling the company from a Singapore accelerator to Tier 1 bank deployments.
Oversees research, development and delivery - the engineering behind Iris and its explainable models.
Three co-founders set out to fight financial crime with AI.
Joined JFDI.Asia in Singapore and raised a seed round led by Wavemaker Partners.
The bank put Silent Eight's AI to work in its sanctions compliance operations.
Raised $6.2M from OTB Ventures, Wavemaker Partners and SC Ventures.
Signed a multi-year deal to embed alert resolution technology in HSBC's global compliance stack.
Closed a round led by TYH Ventures with HSBC Ventures and SC Ventures, reaching ~$55M total.
Grew across Middle East banks and was named a Top 100 Financial Technology Company of 2024.
Launched Iris 7 and won ICA's Compliance AI Solution of the Year 2025 (Europe).
Silent Eight sits in the anti-money-laundering and financial-crime compliance corner of regtech, alongside firms such as NICE Actimize, ComplyAdvantage, Feedzai, Featurespace, Napier, Quantexa, SymphonyAI and Hawk AI. Its wedge in that crowded market is explainability - decisions that come with written, auditable reasoning rather than opaque scores - and a preference for narrow, task-specific models over a single general-purpose one.
The name nods to "888" - eight is a lucky number in many Asian cultures, fitting for a Singapore-founded firm.
Its X/Twitter handle is @whereiseight - literally asking "where is eight?"
Silent Eight deliberately avoided building on a single large language model, training narrow task-specific models instead.
The founding trio had worked together for close to 20 years before scaling the company.