I gave a hands-on lecture on agentic AI at the A*STAR Leadership Retreat, to a distinguished audience comprising A*STAR’s Chief Executive and the two levels of management below him.
For this audience, the important questions extend beyond what an agent can do. Which decisions should we delegate? What information should an agent have access to? Who checks its work, and who remains accountable when it gets something wrong? The session combined practical experience with AI tools and the governance decisions that leaders need to make. No programming required.
What we covered
- The AI landscape and mechanics. How models and agents work, what they cost, and where their limitations matter for an organisation.
- Checking the evidence. Why a polished recommendation can still be wrong, and how to trace its claims back to the original documents.
- Governing agents. Applying the IMDA Model AI Governance Framework for Agentic AI to access limits, permitted actions, human approval, and accountability.
- Governance in practice. An instructor-led example using ATLAS, IMCB’s AI business-intelligence platform, to connect policy choices with technical controls.
The hands-on exercise used a fictional procurement case: five suppliers bidding for a research test platform, with 47 PDFs of requirements, bids, and supporting evidence. Participants first used an agent to understand the case, prepare a procurement brief, and produce a presentation they could check against the sources.
They then designed controls for a procurement workflow: Ingest → Verify → Draft → Review. This made the governance questions concrete. The agent writing the requirements should not read the bids; each verifier should see only its assigned bidder; and consequential decisions need named human owners and clear review points. Participants recorded their chosen controls, residual risks, and proposed tests.
The goal was to give leaders enough direct experience to ask better questions of their teams, and to make informed decisions about where AI belongs in their organisation.
Materials
The AI for Leaders slides and workshop materials are available through OcelliQ’s AI resources. The workshop materials include the fictional procurement case and exercises for preparing a brief, checking evidence, and designing governance controls.
Recording
If you were at the retreat, thank you for joining us! Add me on LinkedIn.