The Vision: From AI User to Corporate AI Entrepreneur (CAIE)

July 2026

Under the Malaysia Digital 2030 (MD2030) action plan, Malaysia is actively pivoting to become an AI Nation by transitioning from mere technology consumers to creators and exporters of "Made by Malaysia" digital solutions. To fulfill this mandate, Dr. Suresh Kumar Krishnan has developed a groundbreaking 5-Day Masterclass titled "The Future of Data: From Analytics to Artificial Intelligence – AI User to Corporate AI Entrepreneur (CAIE)". This curriculum is designed to bridge the gap between national policy and corporate execution by focusing on advanced AI System Thinking.

The program directly supports several MD2030 Strategic Thrusts:

Thrust 4 (Talent): Aims to reskill 700,000 workers and increase the number of skilled workers to 35%. It transitions professionals from passive "AI Users" to Corporate AI Entrepreneurs (CAIE) capable of deploying enterprise-grade AI.

Thrust 2 & 7 (Economy & Innovation): Supports the goal of achieving a 30% digital economy GDP contribution. It trains participants to build Compound AI System Solutions (CAISS) and leverage the AI Agent Value Chain (AIAVC).

Thrust 6 (Trust & Safety): Aligns with the national goal of ensuring 80% of Malaysians trust AI systems. It embeds accountability, transparency, and fairness into corporate AI architectures by teaching AI Governance (AIG) and mitigating AI Bias Risk (AIBR).

Whole-of-Nation Approach (Data as the New Oil): Teaches leaders to apply the Science of Variation & Statistical Thinking, treating data as the foundational asset for the AI Value Chain (AIVC).

---

The 5-Day Masterclass Journey

The curriculum structured by Dr. Suresh progresses from conceptual foundations to hands-on commercialization:

• Day 1 (Foundations of AI and Data Mastery): Demystifies AI, Machine Learning, and Deep Learning; introduces the DRSK Framework for AI system thinking and mitigating AI Bias Risk (AIBR).
• Day 2 (AI Development Lifecycle & Governance): Covers the AI Value Chain (AIVC) from acquisition to monitoring, as well as Levels of Automation (LoA) and Data Mobility (DM).
• Day 3 (Architecting the Future): Focuses on designing Compound AI System Solutions (CAISS) and exploring the four stages of the AI Agent Value Chain (AIAVC): Perception, Cognition, Action, and Learning.
• Day 4 (AI Governance & Future Trends): Deep dives into responsible AI (accountability, transparency, fairness), Explainable AI (XAI), and AI Safety.
• Day 5 (Hands-On Commercialization & Business Modeling): Covers tool selection, cost-analysis, new business model generation, and a final showcase pitching software/app prototypes.

---

The Technical Blueprint: AI Agent Engineering Framework

To build production-grade AI agents, Dr. Suresh Kumar Krishnan outlines a structural framework composed of 17 core engineering disciplines organized into a layered architecture, plus 7 newly identified disciplines required to prevent failures in live deployments.

Rather than a flat list, these disciplines are structured into five build layers sitting on top of a foundation model, wrapped above and below by two comprehensive bands:

1. Layer 0 — Foundation Model: The underlying LLM or multimodal model being wrapped.
2. Layer 1 — Components: What the agent is made of, consisting of Prompt, Context, Memory, Tool Layer, and Data/Knowledge Engineering.
3. Layer 2 — Execution: How a single agent actually runs, including the Harness Engineering (the scaffolding and runtime) and Loop Engineering (the think-act-reflect cycle).
4. Layer 3 — Agent Design Engineering: Defining the agent's role, persona, decision boundaries, and authority level.
5. Layer 4 — Orchestration Engineering & Multi-Agent Systems: How multiple agents coordinate, task-route, and sequence as a single system.
6. Governance Band (Wraps all layers from above): Focuses on safety, compliance, testing, and observability.
7. Operations Band (Wraps all layers from below): Focuses on day-to-day management, model updates, software delivery pipelines, and compute/token cost management (AgentOps, MLOps, DevOps, FinOps).

---

The Aviation Analogy

To make this complex architecture highly intuitive for enterprise leaders, the framework maps each discipline onto its aviation equivalent:

AI Agent Discipline	Aviation Equivalent	Core Concept
Prompt Engineering	The radio script / clearance instructions	The direct instructions given to the pilot.
Context Engineering	Cockpit instruments & situational picture	The immediate information presented in the cockpit at runtime.
Memory Engineering	Flight log & pilot's accumulated experience	The cumulative records and history of the flight/system.
Data/Knowledge Engineering	Pre-flight charts, manuals, & maintenance	The prepared materials and library compiled before takeoff.
Tool Layer Engineering	Aircraft control surfaces, engines, hydraulics	What the pilot physically operates to interact with the environment.
Harness Engineering	The cockpit itself	The interface that wires the pilot, controls, and machine together.
Loop Engineering	Pilot's scan-decide-act cycle (OODA loop)	The continuous iteration cycle of observing, planning, and acting.
Agent Design Engineering	Pilot's role and certification	The defined responsibilities (e.g., Captain vs. First Officer).
Orchestration Engineering	Air traffic control	Real-time coordination and sequencing of multiple aircraft.
Multi-Agent Systems	The fleet	How multiple independent aircraft operate as one unified system.
Safety Engineering	Stall warnings, TCAS, stick shakers	Built-in technical systems designed to prevent failure.
Agent Security / IAM	Cockpit door locks, crew authentication	Restricting access and securing system entry.
Evaluation Engineering	Simulator checkrides and certification	Testing capability in a safe, adversarial environment before flying.
Human-in-the-Loop Oversight	Two certified pilots & ground override	Collaborative decision gates and manual override authority.
Alignment & Policy Engineering	ICAO regulations and airline SOPs	Standard rules ensuring behavior matches organizational intent.
Observability Engineering	Flight data recorder / black box	Logging and monitoring tools to reconstruct what happened and why.
AgentOps	Airline daily operations control center	Live operational monitoring and health management of active agents.
MLOps	Aircraft maintenance & avionics upgrades	Upgrading and maintaining the underlying core model engine.
DevOps	Software-update pipeline for avionics	Delivering updates to the software scaffolding surrounding the model.
FinOps	Fuel management and route economics	Governance over token spend and running costs.
Incident Response / BCP	Emergency checklists & diversion protocols	Rollback and kill switch execution when an agent fails in production.
Agent UX / Product Engineering	Passenger-facing cabin experience	Designing how the end-user interacts with the service.
Knowledge Distillation	Incident learnings compiled back into manuals	Feeding real-world learnings back into the core knowledge base.

---

The Seven Critical Extensions

The framework emphasizes seven disciplines that standard enterprise blueprints often omit, but which are essential for production-grade deployment:

1. Data / Knowledge Engineering: The upstream pipeline that cleans, chunks, and embeds the knowledge base.
2. Evaluation Engineering (Evals): Robust pre-deployment adversarial testing (such as golden datasets and red-teaming).
3. Agent Security / IAM: Managing credentials, sandboxing, and least-privilege tool access.
4. Human-in-the-Loop Oversight: Explicitly designing escalation pathways and approval gates rather than assuming them.
5. Compliance (Split from Alignment): Separating adherence to external laws (e.g., PDPA, Bank Negara Malaysia (BNM), and MCMC) from internal corporate intent.
6. Agent UX / Product Engineering: Building user trust and addressing human interface friction, which is the most common bottleneck to adoption.
7. Incident Response / Business Continuity: Creating concrete runbooks, rollback plans, and emergency kill switches for production failures.