The Challenge
Temasek Trust Asset Management is the asset management arm of Temasek Trust, a MAS CMS-licensed manager running a multi-asset, private markets, and impact portfolio across two investment units. As a lean allocator that deploys much of its capital through external managers, its returns exist to sustain philanthropic disbursements and deliver measurable impact, which sets a distinctive context for how AI tools should and should not be used.
The defining challenge was the bar set by the room. The workshop was for the entire team across investment, operations, and supporting functions, but senior management, including the CEO, were present and were themselves confident, well-informed users of Generative AI. That made a standard awareness session insufficient. The programme had to be useful and accessible to the full team while still engaging the most AI-fluent people in the room, which called for a trainer who could meet power users as a genuine practitioner rather than work through introductory slides.
Why Heicoders Academy
TTAM chose Heicoders Academy for two reasons that mattered to this particular audience. The first is that our instructors are active practitioners, not career trainers, which is what allows us to hold a credible conversation with a team that already knows the basics and wants depth. The second is the extent to which we customise. Rather than reuse a generic deck, we re-anchored every example to TTAM's real investment work, re-built the curriculum around the platform TTAM actually uses, and carried a financial-institution-appropriate governance lens through every section.
Given the profile of the audience, and the senior stakeholders in particular, we made a deliberate decision about who should lead. We sent Beh Min Yan, our CEO and Principal Instructor, an ex-Machine Learning Engineer at GoJek and ex-Quantitative Researcher at Credit Suisse, to run the session personally. It was a signal of how much we prioritise this partnership, and it ensured even the most experienced people in the room were engaged by someone who could meet them on their own level.
The Solution
We designed and delivered a customised workshop built entirely around Claude as the primary tool, spanning the claude.ai chat interface, Research mode, Claude Cowork, and Claude Code. Every example was re-anchored to TTAM's own mandate, with a single throughline running across the day: one prompting toolkit, recombined across document types and tools, with verification treated as a constant. The curriculum covered:
- Prompt engineering foundations: Shot-based prompting, prompt chaining, and system prompting, taught through TTAM-real examples such as a "Quarterly Review to Trustee Update" chain and an investment-operations-analyst persona, brought together in a complete Task, Context, Reference, Evaluate, Iterate framework.
- Reliability and judgment: How and why AI hallucinations occur, and how to reduce them through data grounding, mandatory citations, and negative constraints, with verification of every extracted figure against source documents presented as a non-negotiable discipline for a regulated institution.
- Applied use cases: Using Research mode to turn cross-source web investigation into a properly cited brief, and drafting professional emails, investment research reports, and compliance-ready documentation, each mapped back to the prompting technique it relies on.
- Hands-on investment workflows and agentic automation: Building populated investment monitoring decks and tracking spreadsheets from real public company filings, analysing portfolio performance data into summary reports, and progressing into agentic automation with Claude Cowork across a set of investment-relevant workflows.
- Advanced features and choosing the right tool: Skills, Connectors, Projects, Artefacts, and Claude Code, closing with a clear map of when to use Chat, Research mode, Cowork, Code, or a custom agent, framed for a financial institution as a spectrum of autonomy, setup, and governance.
A consistent theme was honesty about fit. Where a tool or example was not the right one — such as accuracy-critical reconciliation work, certain limits of Research mode, or full-stack web app builds — we said so plainly, which is part of how we build trust with an expert audience.
Results
The programme met its central test: it served the full team while holding the attention and respect of the senior, AI-fluent stakeholders in the room. By grounding every module in TTAM's actual investment work and using real public filings for the worked examples, the session moved participants from general familiarity to a sharper, more disciplined command of Claude across their day-to-day tasks, with verification and governance built into how they work rather than bolted on afterwards.
What's Next
The engagement reflects a growing appetite among Singapore's asset management and financial services firms for AI training that goes well beyond awareness and is tailored to the realities of a regulated, mandate-driven environment. We remain a committed partner to TTAM, ready to support deeper runs, a dedicated data-governance and safe-use segment for the regulated context, and broader adoption across investment, operations, and impact functions as the team's AI capability continues to mature.
