Singapore’s wealth management industry is a global powerhouse, set to manage a projected US$239 billion for high net worth individuals (HNWIs). Yet a significant portion of this wealth, roughly US$60 billion, remains untapped and sitting idle in cash and cash equivalents.
This is the cost of a fundamental disconnect. The very hours you, as a successful individual, spend earning your money (the 9-to-5) are the same hours your financial advisor is available. The crucial window of your time, from 5-to-9, when you finally have the headspace to think about your financial future, is often met with an “out of office” reply.
This friction leads to decision paralysis. Questions go unanswered, opportunities are missed, and money fails to compound effectively. This article explores this paradox and introduces agentic AI as a potential approach. It also examines foundational changes firms would need to make to deliver financial guidance on a 24/7 basis.
The US$60 billion bottleneck
The scale of Singapore’s wealth is staggering, with the nation attracting a net gain of approximately 3,500 HNWIs in 2024 alone. The advisory model, however, has not kept pace. With around 20,000 financial advisors serving the growing population, a traditional, meeting-based structure creates inevitable bottlenecks.
The result is a significant pool of latent capital. The global average for HNWI cash holdings is around 25% of portfolios. Applied to Singapore’s US$239 billion HNWI market, this translates to nearly US$60 billion in assets that are not working as hard as their owners. While a cash buffer is necessary, much of this remains uninvested, stalled by indecision. Industry research and practitioner observations have consistently pointed to a lack of know-how as a key barrier to investment in new asset classes, a challenge exacerbated by the 9-to-5 advice gap.
Let’s imagine a different reality where a financial co-pilot operates on customers’ schedules. This co-pilot scans news feeds in the evenings, picks up material signals, and alerts you with relevant context. You engage with it by analysing the agent-influenced impact on your portfolio and the proactive adjustment it suggests before making final decisions.
Beyond chatbots: Engineering customers’ proactive AI co-pilot
The solution described above is agentic AI, a proactive, goal-oriented system. Unlike a chatbot that reacts to commands, an agent perceives its environment (market data, customers’ portfolios, and stated goals), reasons through complex information, and takes autonomous action on the customer’s behalf, with appropriate permissions.
While the prospects are promising, many firms struggle to scale AI beyond proofs of concept and deliver value through production-grade assets. A commonly referenced approach is a six-pillar framework intended to establish foundations for enterprise scaling:
- Foundational architecture: A robust, scalable technical architecture upon which AI workloads and agentic systems can be built.
- Operating model: Organisational processes and team structures that support delivery of business value, with AI treated as a core component.
- Readiness of data: Open yet secure access to high-quality data for model training and inference.
- Experience for humans and AI: User-centric design that enables AI to augment human capabilities rather than introduce friction or distrust.
- Strategic alignment: Clear alignment between AI initiatives and broader business strategy to support measurable outcomes.
- Trusted AI: Governance, transparency, and ethics embedded from the outset to support responsible scaling.
Let’s take a deeper look at the characteristics of foundational architectures that support AI at scale.
Foundational architectures suitable for agentic AI
A 2025 MACH Alliance Global Annual Research Report indicated that organisations with established composable architectures were more likely to report broader AI adoption and positive outcomes than those early in their journey, suggesting a link between modular technical foundations and the ability to scale AI initiatives.
Modularising and decoupling existing technology can allow open yet secure access to underlying data through APIs and model context protocol (MCP). MCP is an open standard that enables agents to connect to data sources and tools through a standardised interface. This allows AI models to query and interact securely with business services to perform complex tasks, including actions taken on behalf of users.
For example, an autonomous agent responding to a market event could use MCP to access portfolio details via a client API, calculate revised allocations based on market intelligence, and send those revisions to pricing systems for scenario analysis. The agent could then present a ranked set of options for user selection.
Additionally, modularity and fine-grained access to APIs and data can support a zero-trust strategy, where agents are granted access only to the APIs they require. This is particularly important in wealth management, given security and privacy requirements enforced by MAS (Monetary Authority of Singapore).
Another architectural consideration is evaluation. Effective deployment requires frameworks and tooling that support both online and offline evaluation. Evaluations are systematic assessments used to measure an agent’s performance and reliability, ensuring tasks are executed accurately. LangChain’s State of Agent Engineering report shows that only 38.3% of firms currently evaluate performance using production data.
The point is straightforward. Agentic initiatives should not be treated solely as technology projects. Scaling AI effectively requires architectural decisions that are reinforced by operating models, governance, and user experience.
The future of wealth is 24/7
Wealth management increasingly centres on augmenting financial advisors with intelligent tools that help clients make more timely decisions. A human-plus-machine approach could help address the 5-to-9 advice gap and reduce the amount of capital left idle.
The technology to support this approach exists. Whether it delivers meaningful outcomes depends on how deliberately firms integrate technical, organisational, and governance considerations into their AI strategies.
