FA-0601Agentic & Generative AIAI for Leaders & Business

Demystifying AI: From Basics to Agents

A one-day conceptual overview of models, tools and responsible agent design

Understand how modern AI models and tool-using agents fit together, then review a responsible use-case proposal.

Introduction

Why this course

This one-day overview explains AI foundations, machine learning, deep learning, language models and tool-using agents at a conceptual level. It connects prompting and MCP integration to business use cases, limitations and governance questions.

The emphasis is on informed discussion and a small responsible-agent proposal, not training models or deploying a production agent in one day. Practical examples are illustrative and do not imply client relationships, verified adoption rates or autonomous reliability.

Learning outcomes

The course teaches participants to:

Distinguish AI capability labels, model-learning approaches and agent behaviour.
Explain training and evaluation concepts and the roles of neural networks and language models.
Frame prompts with clear context, constraints and review criteria.
Describe MCP host/client/server roles and the difference between tool access and application orchestration.
Analyse an agent use case with permissions, human oversight and failure handling.
Discuss bias, privacy, accountability and the limits of explainability.

Prerequisites

Basic familiarity with AI/ML terminology (e.g., “model,” “training,” “dataset”)
Comfort with high-level concepts in programming, statistics, or data processing
Openness to discuss real-world examples and ethical dilemmas

Training outline

7 modules

01Module 1 — AI Foundations2 topics

Reactive vs. Deliberative vs. Learning agents
Narrow AI vs. Artificial General Intelligence (AGI)

Distinguish agentic behaviour from model capability and deployment marketing; AGI is a concept, not an assumed capability of a current product.

02Module 2 — Machine Learning and Deep Learning3 topics

Overview: supervised, unsupervised, reinforcement learning
Deep neural networks, architectures (CNNs, RNNs), backpropagation
Use of Deep Learning in agentic systems

Training, validation, evaluation and data limitations through high-level examples rather than a full implementation lab.

03Module 3 — Language Models and Prompting1 topics

What makes LLMs 'large': data, transformer architecture, token prediction

Context windows, instructions, examples and iterative review; distinguish plausible text generation from verified knowledge.

Reasoning and chain-of-thought prompting as model-dependent concepts, not universal requirements or evidence of correctness.

Structured task requests and source checking before tools or downstream actions.

04Module 4 — MCP and Tool Integration4 topics

MCP host, client and server roles; tools, resources and prompts as supported primitives.
Protocol/version and capability compatibility; tool access remains subject to application permissions.
MCP exchanges context and tool capabilities; the AI application manages its model use and orchestration.
Data permissions, untrusted tool content and prompt-injection risks; protocol use does not itself establish safety.

05Module 5 — Agentic Systems and Use Cases4 topics

Systems that use models to plan/select actions and interact with tools within defined boundaries.
Autonomy, planning, context and escalation; distinguish automation from reliable independent judgement.
Illustrative customer-service, finance, supply-chain and cybersecurity support scenarios.
Evaluate business value, uncertainty, unauthorised actions and accountability without unsupported adoption or incident statistics.

06Module 6 — Ethics and Governance5 topics

Key principles: fairness, transparency, privacy, accountability
Machine ethics taxonomy: impact, implicit, explicit, full agents
Explainable AI (XAI): transparency in black‑box systems
Governance issues in agentic AI: oversight, human‑in‑the‑loop, audit tracing
Broader risks: job displacement, misinformation, alignment with human values

Explainability can aid understanding but does not prove fairness, safety or accountability; use human ownership and evidence-based review.

07Module 7 — Synthesis and Discussion2 topics

How ML/DL, LLMs, MCP, and ethics combine in end‑to‑end agent systems
Group activity: draft a responsible agent proposal (use case, prompt, tools, safe‑guards)

Discuss deployment and monitoring questions, remaining evidence and how to revise the proposal as technology changes.

A programme built around your team.

Share your training goals and requirements.