Edvance International Holdings (HKEX: 1410), through its Singapore subsidiary Paradise Matrix, and CWG Innovations have launched RealRelay Fintech, the third vertical on RealRelay.ai — a no-code platform for building, testing and running AI agents. The vertical applies agentic AI to finance, beginning with fraud detection, reconciliation and document checks: slow, manual, error-prone work the platform aims to do accurately and in hours, with every step recorded. The launch took place at a youth AI event in Singapore, where participants used agentic AI on a simulated fraud case.
The platform’s structure is the point. Every agent on RealRelay passes through a “Trust and Control Layer” — the platform’s own security agent tests an agent before it goes live and re-tests it whenever it changes. The platform controls permissions, logging, retention and access centrally, and each agent carries a risk rating covering the data it can touch, the actions it can take, and how it behaved under test. The launch covers four areas: research and analysis, decision documents (credit memos, investment committee packs), fraud and compliance checks, and a learning track. Institutional deployments, including Singapore-hosted and on-premises options, are available on request.
The AI risk management deadline
The timing is not incidental. On 7 October 2026, the Monetary Authority of Singapore issued its Guidelines on AI Risk Management, setting supervisory expectations across oversight, AI risk management systems, life-cycle controls and capability. They apply to all financial institutions and all forms of AI, explicitly including “newer developments such as AI agents.” They take effect on 7 October 2027 for the oversight and risk-management sections, and by 7 October 2028 for life-cycle controls. MAS will consult the sector in 2027 on what further agentic-AI guidance would be useful.
The AI risk management guidelines ask for the artefacts a governance platform would generate: an accurate inventory of AI use, a risk materiality assessment weighing impact, complexity and reliance, and life-cycle controls covering testing, human oversight, third-party risk and monitoring. MAS’s industry work has produced an AI Risk Management Toolkit under Project MindForge and a framework for agent runtime safeguards, signalling the regulator is thinking about agent-level controls. A platform that produces an agent inventory, a risk rating and a full log is, in effect, selling the compliance record the deadline demands.
Singapore is also funding adoption. FSTI 4.0 commits S$220 million over three years, with an AI Pathfinder track that co-funds financial institutions adopting market-tested solutions listed on PathFin.ai — up to 50% of costs, capped at S$200,000 for Singapore-based fintechs. PathFin.ai’s participating organisations include HSBC, Standard Chartered and MUFG. CWG says it is preparing to list RealRelay on PathFin.ai, putting it in front of institutional buyers and into the funding track.
What the platform claims
RealRelay is broader than fintech. Its Token Exchange, launched in September 2026, routes tasks across more than 300 AI models under one subscription; the Agentic Layer lets users assemble and deploy agents without code. Education and cybersecurity were the first two verticals. Edvance brings more than 20 years in cybersecurity, lending the “trust-first” positioning credibility — it is, at least, a security business, not a newcomer.
The claims that matter are unproven. The company names no financial institution as a customer, no audit has verified the Trust and Control Layer, and the platform’s promise that an agent’s test record maps to the MAS guidelines is a plan for 2027, not a delivered feature. The youth-event framing — students tracing a fraud in hours — is a demonstration, not institutional deployment.
What to watch
Three things will show whether the vertical matters. First, whether RealRelay appears on PathFin.ai, since that is the route to the AI Pathfinder funding and to institutional buyers. Second, whether a named financial institution adopts it, which would be the first real test of the governance claims. Third, whether the mapping to the MAS guidelines survives the regulator’s promised 2027 consultation on agentic AI, since a moving target can leave a compliance story behind.
What the launch does and doesn’t establish
It establishes a third vertical on an existing no-code agent platform, a trust-and-control layer with testing and risk ratings, and alignment with a 2027 regulatory framework. It does not establish any institutional customer, independent validation of the controls, or that the platform’s artefacts will satisfy MAS’s expectations.
The honest summary: a compliance-first bet, and the compliance is the interesting part. Rather than sell raw agent capability, RealRelay sells the record that Singapore’s financial institutions will soon be required to keep — inventory, risk rating, logs — around a platform that also does the work. That is a sensible response to a real deadline. Whether it becomes a business depends on adoption the launch does not show, and on a framework MAS is still refining.
Editor’s Note: This article draws on the Edvance/CWG release and MAS’s published Guidelines on AI Risk Management and FSTI 4.0 materials, plus independent coverage by The Business Times, CNA, Asian Banking & Finance and Finance Magnates. The platform’s features are as described by its developers; the regulatory dates and framework are from MAS; no customer or outcome data was disclosed. No interview was conducted.