Advanced Agentic AI: Commanding Agent Teams for Unrivaled Productivity
About This Course
This 4-day advanced technical practitioner programme teaches developers, AI engineers and technical architects to engineer, command and govern AI workforces — intelligent reasoning systems that integrate, make sense of, and act upon heterogeneous information sources under human oversight. Day 1 covers agent runtime foundations: the assistant, agent, workflow and agent-team taxonomy; the trigger–route–execute–verify–learn loop; model routing across cloud and local execution; and security architecture and controls for a governed agent sandbox. Day 2 connects agents to enterprise business systems — email, calendar, records, ticketing, databases, documents — through an orchestration fabric with human-in-the-loop approval gates, and assesses emerging standardised tool-access protocols and vendor claims. Day 3 builds persistent, always-on agents with scheduled work and accumulated skills, and establishes data governance standards for agent memory: access, usage, storage, archival and retention across the data lifecycle. Day 4 covers multi-agent Command Centre design; corporate security policies and security governance runbooks aligned with IMDA's Model AI Governance Framework for Agentic AI; and a capstone in which teams engineer and defend a complete AI-native organisation design. Every module teaches a transferable pattern illustrated with a current reference tool, refreshed each cohort; the design decision is assessed, never any product's configuration.
What You'll Learn
• LO1: Analyse the business drivers and application areas of intelligent reasoning systems by classifying AI assistants, agents, workflows and agent teams via the operating loop.
• LO2: Design a model-routing strategy for intelligent reasoning systems across cloud and local execution, justified by cost, latency, quality and data boundaries.
• LO3: Design security architecture and controls for an agent sandbox: define the security specifications of system components, permission boundaries, approval gates and audit logging.
• LO4: Build agent reasoning systems that integrate, make sense of, and act upon heterogeneous information sources across business systems, with human-in-the-loop checkpoints.
• LO5: Assess the potential of emerging technologies, including standardised MCP-style tool-access protocols, evaluating vendor claims and adoption considerations.
• LO6: Configure a persistent agent and develop internal standards for data governance of its memory: access, usage, storage, archival and retention across the data lifecycle.
• LO7: Establish benchmarks and evaluate deployed agents: task success, hallucination checks, tool-use correctness, escalation rate, cost, regression, red-teaming.
• LO8: Design a multi-agent Command Centre with role design, goals, budgets, permission scoping and monitoring, and judge when a multi-agent system is not the right answer.
• LO9: Develop corporate security policies, security governance frameworks and incident-handling runbooks for an agent workforce, aligned with IMDA's Model AI Governance Framework for Agentic AI.
• LO10: Engineer an AI-native organisation design for one business function, applying security architecture, data governance and intelligent reasoning decisions, defended to a technical panel.
Entry Requirements
Participants are expected to have:
• Comfort with APIs and general automation concepts
• Comfort with technical setup and self-hosted/local tooling
• Basic generative AI familiarity
• Willingness to work hands-on with local and self-hosted tools
• Basic scripting knowledge is helpful but not mandatory