AI Agents AI Governance AI Academy About Connect →
AI Academy — Practical notes. No consultancy pitch.

Practical AI.
Plain English.

A small library of guides shaped by enterprise work in AI agents, governance, and Microsoft 365. Free to read, free to keep.

All courses free No technical background needed Built for enterprise professionals
— COURSE 001
Free

AI Agents, Decoded.

Five plain-English chapters for anyone ready to understand what's actually happening inside the AI revolution. How agents work, where they fail, and how to control them.

Your progress 0 of 5 complete
01
Foundation

Chatbot, Copilot, Agent — what's actually different

Three terms used interchangeably that mean three very different things.

+
  • Why a chatbot is not an agent, and why the difference matters
  • What Copilot actually does, and what it can't do alone
  • The defining traits of a true autonomous agent
  • A mental model you'll use every time someone pitches you "AI"
"An agent doesn't just answer. It decides, acts, and keeps going — often without asking for permission."
02
Mechanics

How an agent thinks and acts — no code required

Perception, reasoning, memory, tools. The four-part engine behind every agent worth knowing about.

+
  • The perceive, reason, act loop — explained without jargon
  • What "memory" means for an AI, and its surprising limits
  • Tools: how agents connect to calendars, data, APIs, the real world
  • Multi-agent systems: when agents hand tasks to other agents
"AI doesn't remove responsibility. It redistributes it."
03
Risk

Where agents break — the risks nobody talks about

Hallucinations are just the beginning. A candid look at what makes enterprise AI deployments fail quietly.

+
  • Hallucination, drift, and silent failure: a practical taxonomy
  • Why "it worked in the demo" is one of the most dangerous sentences in AI
  • The accountability gap: when something goes wrong, who owns it?
  • Real failure patterns from real enterprise deployments
"The risk isn't that the agent does something dramatic. It's that it does something quietly wrong, repeatedly, at scale."
04
Governance

What governance looks like before you deploy one

The questions every organisation must answer before a single agent touches a live system.

+
  • Ownership, oversight, and audit trails: the governance triad
  • EU AI Act, ISO 42001, and what they actually require from you
  • The 7-stage agent lifecycle: from ideation to decommission
  • A pre-deployment checklist you can use on Monday
"Governance isn't bureaucracy. It's the difference between an agent that earns trust and one that quietly erodes it."
05
Next steps

Your first agent — questions to ask before you build

The chapter that turns understanding into action. A decision framework for the room where it gets decided.

+
  • The problem-fit test: not every problem needs an agent
  • Build vs buy vs configure: the honest breakdown
  • 10 questions to ask your business stakeholder before you commit
  • What a well-governed first agent looks like in practice
"The best first agent isn't the most impressive one. It's the one with the clearest boundaries, the simplest task, and a human watching closely."
— COURSE 002
Free

AI Agent Governance, in Plain English.

Six chapters on what governance actually means, who is accountable, how the 7-stage lifecycle works, what the EU AI Act requires, and how to build a programme that works on Monday. No consultant required.

6 chapters ~50 min read Builds on Course 001 RACI framework included
Your progress 0 of 6 complete
01
Definition

What is AI governance, and why does it matter now

Not a compliance checkbox. A practical look at what governance actually is, and why most organisations are already behind.

+
  • What governance means in the context of AI agents specifically
  • The three things that go wrong without it: accountability gaps, audit failures, and trust erosion
  • Why the AI Act has made this urgent, not optional
  • A one-sentence definition your board will actually understand
"Governance is what separates an agent that's deployed from an agent that's trusted."
02
Foundation

Why governance fails before it starts

The most common governance failures happen long before an agent goes live. Here is what to watch for.

+
  • The "we'll deal with it later" trap — and why later never comes
  • Governance by committee: why too many owners means no owner
  • Documentation that exists but never gets read
  • How to spot a governance programme that is theatre, not substance
"A governance framework nobody follows is not a safety net. It is a liability."
03
Structure

The people behind the agent

Every agent needs humans attached to it. The 14-role RACI framework, explained in plain English.

+
  • RACI in 60 seconds: responsible, accountable, consulted, informed
  • The 14 roles that touch an AI agent across its lifetime
  • Who signs off on deployment, and who carries the risk if it goes wrong
  • How to map this to your existing team structure
"The question is not who built the agent. The question is who owns it on a Tuesday afternoon when something breaks."
04
Lifecycle

From idea to decommission

The 7-stage agent lifecycle model. Where governance checkpoints sit, and why each one matters.

+
  • The 7 stages: ideation, design, build, test, deploy, monitor, decommission
  • What a governance gate looks like at each stage
  • Why decommissioning is the most neglected stage in enterprise AI
  • A one-page lifecycle map you can share with stakeholders
"Most organisations think about agents at the build stage. Governance starts at the idea stage."
05
Compliance

What the regulations actually require

EU AI Act, ISO 42001, NIST RMF. What each framework demands from you in plain terms.

+
  • EU AI Act: risk tiers, prohibited uses, and what high-risk means for your organisation
  • ISO 42001: a management system for AI, not a technical standard
  • NIST RMF: the American framework gaining traction in global enterprises
  • A side-by-side comparison: what maps across, what doesn't
"You don't need to read 300 pages of regulation. You need to know which 10 requirements apply to your situation."
06
Action

Building your governance programme

The chapter that moves from theory to Monday morning. A practical starting point for any organisation, regardless of size.

+
  • The five components every governance programme needs from day one
  • How to audit an existing agent for governance gaps in under an hour
  • A starter template: policy, owner, review cycle, escalation path
  • What good looks like at 30 days, 90 days, and 12 months
"You do not need a perfect programme. You need a real one: documented, owned, and reviewed on a schedule."
— COMING NEXT

More courses in the pipeline.

COURSE 003
Microsoft Copilot:
What Leaders Need to Know
COURSE 004
More to come

Who this is for

  • Business leaders evaluating AI adoption
  • Team leads asked to "do something with AI"
  • Governance and risk professionals
  • Anyone curious — no background needed
Part of the Athar Sheikh AI Academy — practical notes for enterprise professionals navigating the AI era.
Based in Lausanne, Switzerland.
This site is a personal profile and knowledge space, not a consultancy offering. It reflects practical work and frameworks in enterprise AI governance and Microsoft 365 Copilot.
Personal profile Employer neutral AI governance AI Academy