Exam Prep

EXIN AI Foundation

Exam Preparation for IT Professionals

You know AI. Let's nail the exam.

Exam At A Glance

40
Multiple-choice questions
60 min
Time limit -- 90 seconds per question
65%
Pass mark -- 26 out of 40
Bloom L1-L2
Remember + Understand (no analysis)
No Open Book
No notes, no materials allowed
No Devices
No electronic equipment or aides

Topic Weights

1. Intro to AI & History15%
2. Ethics & Legal15%
3. Enablers of AI15%
4. Finding & Using Data20%
5. AI in Your Organization20%
6. Future Planning15%
Topics 4 + 5 = 40% combined -- your biggest opportunity
🎯

Key Insight

The exam is ~50% governance, ethics & business -- NOT technical AI.

That's where most candidates lose marks. Don't over-study neural networks and under-study stakeholder grids.

Course Agenda

Module 0
Course Introduction 5 min
Module 1
AI History & Definitions Normal
Module 2
Ethics & Legal Slow
Module 3
Enablers of AI Fast
Module 4
Finding & Using Data Normal
Module 5
AI in Your Organization Slow
Module 6
Future & Impact Normal
Module 7
Exam Prep & Review Fast
Topic 1
15%

Introduction to AI & History

Pace: Normal

"The mental quality that consists of the abilities to learn from experience, adapt to new situations, understand and handle abstract concepts, and use knowledge to manipulate one's environment."
-- Definition of Human Intelligence (EXIN syllabus)
Exam Tip

This exact definition appears in exam questions. Note: it's about HUMAN intelligence, not AI.

Core Definitions

Artificial Intelligence (AI)
"Intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals."
Machine Learning (ML)
"The study of computer algorithms that allow computer programs to automatically improve through experience."
Scientific Method
"An empirical method for acquiring knowledge that has characterized the development of science."
Hierarchy

AI is the umbrella. ML is the approach. DL is the technique.

🔬

Scientific Method

Scientific Method = EMPIRICAL method for acquiring knowledge

Exam Trap

The key word is EMPIRICAL -- based on observation and experiment, not theory. The exam offers distractors like "procedural" or "technological" approach.

Exam Question

What is the scientific method?

A) A procedural approach to solving problems
B) A technological approach to processing data
C) A teaching method used in universities
D) An empirical method for acquiring knowledge
Answer: D
The scientific method is defined as "an empirical method for acquiring knowledge." The key word is EMPIRICAL. Options A, B, and C sound plausible but miss this critical term.

AI Timeline -- Key Milestones

1956
Dartmouth Conference -- AI coined as a field
1974-1980
First AI Winter -- unmet expectations, funding cuts
1987-1993
Second AI Winter -- expert system limitations
2010s
Big Data + IoT revolution fuels AI resurgence
2017
Asilomar AI Principles published (FLI)
2022
LLMs go mainstream -- ChatGPT, public AI interest explodes
🏫

Dartmouth Conference, 1956

Four researchers coined "Artificial Intelligence" as a field of study.

  • John McCarthy
  • Marvin Minsky
  • Nathaniel Rochester
  • Claude Shannon
Fun Fact

The original proposal claimed AI could be "solved in one summer." It's been nearly 70 years.

The Two AI Winters

First Winter: 1974-1980

  • Overblown promises from Dartmouth era
  • Unmet expectations
  • Government funding cuts
  • Computing power insufficient
vs

Second Winter: 1987-1993

  • Expert systems proved too expensive
  • Narrow applicability
  • Maintenance nightmare
  • Desktop PCs outperformed specialized hardware
Memory Trick

Winter 1 = Seventies Slump (promises > reality) | Winter 2 = Eighties Expert Failure (expert systems flopped)

Asilomar AI Principles (2017)

23 Principles in 3 Categories
  1. Research Issues -- responsible research goals
  2. Ethics & Values -- safety, transparency, human values
  3. Longer-term Issues -- superintelligence, existential risk

Published by: Future of Life Institute (FLI), 2017

Purpose

"To ensure AI is developed safely and beneficially"

Exam Tip

Asilomar = guidelines for responsible AI development. They are VOLUNTARY, not law.

Exam Question

What is the purpose of the Asilomar AI principles?

A) To enforce legal penalties for misuse of AI
B) To create technical standards for AI systems
C) To ensure AI is developed safely and beneficially
D) To establish AI certification programs
Answer: C
Asilomar principles are voluntary guidelines to ensure AI is developed safely and beneficially. They don't enforce laws (A), set technical standards (B), or create certification programs (D).

Types of AI

Narrow / Weak AI (ANI)

  • Task-specific, well-defined domains
  • Exists today
  • Image recognition
  • Speech-to-text
  • Language translation
  • Virtual assistants (Siri, Alexa)
  • Spam filtering
vs

General / Strong AI (AGI)

  • Any intellectual task a human can do
  • Hypothetical -- does NOT exist yet
  • Would learn, reason, plan across domains
  • Self-aware understanding
Common Misconception

ChatGPT feels like AGI but is still narrow AI -- it can't drive a car or do your taxes.

Exam Question

Which type of AI performs specific tasks within a defined scope?

A) General AI
B) Domain AI
C) Narrow AI
D) Universal AI
Answer: C
Narrow AI (also called Weak AI or ANI) is task-specific and operates within a defined scope. "Domain AI" (B) sounds right but is NOT an official EXIN term -- classic trap.

Ethical Principles for AI

Floridi & Cowls (5 Principles)

Beneficence
Do good -- promote well-being
Non-maleficence
Do no harm -- avoid damage
Autonomy
Preserve human decision-making power
Justice
Fair and equitable treatment
Explicability
Transparency -- explain AI decisions

UK AI Principles (5 Principles)

Safety, Security & Robustness
AI must function safely and securely
Transparency & Explainability
How and why AI makes decisions
Fairness
No discrimination or bias
Accountability & Governance
Clear responsibility for AI systems
Contestability & Redress
Challenge and correct AI decisions
🧠

Memorize These Frameworks

Floridi & Cowls

Beneficence Non-maleficence Autonomy Justice Explicability

"Be Nice And Just Explain"

UK AI Principles

Safety & Security Transparency Fairness Accountability Contestability

"STFAC"

Next: SORT EXERCISE -- students sort principles into two frameworks
SORT EXERCISE

Sort: Floridi & Cowls vs UK AI Principles

Which framework does each principle belong to?

Safety
Beneficence
Fairness
Autonomy
Transparency
Non-maleficence
Accountability
Justice
Contestability
Explicability
Floridi & Cowls (BNAJE)
  • Beneficence
  • Non-maleficence
  • Autonomy
  • Justice
  • Explicability
UK AI Principles (STFAC)
  • Safety
  • Transparency
  • Fairness
  • Accountability
  • Contestability

AI's Impact on Society

🌐 Social Impact
Job transformation, communication changes, new skills needed, ways of working
💰 Economic Impact
Automation of industries, job displacement AND creation, market volatility
🌱 Environmental Impact
Energy consumption, water usage for cooling, carbon footprint of data centers
🌎 UN SDGs
17 Sustainable Development Goals -- AI can help or hinder progress
🏛 EU AI Act (2024)
First comprehensive AI legislation, risk-based classification

Exam Question

Which is an ENVIRONMENTAL impact of AI?

A) Demand for new skills in the workforce
B) Increased cost of AI implementation
C) Changes in communication patterns
D) Water usage demand for cooling data centers
Answer: D
Water usage for cooling is environmental. New skills (A) = social. Cost (B) = economic. Communication (C) = social. The exam tests whether you can categorize impacts correctly.

Sustainability Measures for AI

🌿 Green IT
Renewable energy, efficient hardware, green data centers
⚡ Data Center Efficiency
Optimize cooling, power usage effectiveness (PUE), location selection
📦 Sustainable Supply Chain
Responsible sourcing, circular economy for hardware
🧮 Algorithm Choice
Select energy-efficient models, avoid over-engineering
💻 Low-Code / No-Code
Reduce computational overhead with simpler solutions
📊 Monitoring & Reporting
Track and report environmental impact continuously

Exam Question

What helps limit the environmental impact of AI?

A) Using more complex AI models for better accuracy
B) Using green, energy-efficient AI models
C) Increasing the number of data centers
D) Extending AI training times for better results
Answer: B
Green, energy-efficient models reduce environmental impact. More complex models (A) and longer training (D) INCREASE impact. More data centers (C) also increases energy use.
Topic 2
15%

Ethics & Legal Considerations

Pace: SLOW -- high exam value

⚖

What Is Ethics?

Ethics = Moral principles that govern a person's behavior or the conducting of an activity

In AI: guidelines that govern the creation, development, and use of AI systems

Ethics vs Law

Ethics

  • Voluntary guidelines
  • Moral, self-imposed
  • No legal penalties
  • Changes with society
  • Industry self-regulation

Example: Asilomar Principles

vs

Law

  • Mandatory compliance
  • Enforceable, penalties
  • Fines, prosecution
  • Codified regulations
  • Government enforced

Example: GDPR (fines up to 4% revenue)

Remember

GDPR = LAW (fines). Asilomar = ETHICS (voluntary). The exam tests this distinction.

Key Ethical Concerns in AI

⚠ Bias & Discrimination
Training data reflects historical prejudices; AI amplifies them
🔒 Data Privacy
Collection, storage, and use of personal data without consent
💼 Employment Impact
Job displacement, changing skill requirements, economic disruption
💣 Autonomous Weapons
AI-controlled weapons with lethal decision-making capability
🚗 Autonomous Vehicles
Liability framework: who is responsible when an autonomous vehicle causes harm?
Real World

Amazon's hiring AI discriminated against women -- it learned bias from historical training data that favored male candidates.

Next: SPOT THE MISTAKES -- pairs find 3 errors in a scenario
SPOT THE MISTAKES

Find the 3 Errors

Work in pairs. This AI project description has 3 factual errors. Can you find them all?

"TechCorp implemented an AI-powered recruitment tool. To ensure quality, they certified their AI management system under ISO 9001. During the risk assessment phase, the team applied Crisis analysis to identify potential threats. The project manager was designated as the risk owner, ultimately accountable for ensuring all risks were managed appropriately."

ISO 9001 ISO 42001 -- ISO 42001 is for AI Management Systems (AIMS).

Crisis analysis Risk analysis / SWOT / PESTLE / Cynefin -- Crisis is NOT a technique.

project manager the individual ultimately accountable -- Risk owner is the accountable individual, not necessarily the PM.

Exam Question

What is 'ethics in AI'?

A) Regulations enforcing AI safety standards
B) Guidelines governing the creation and use of AI
C) Technology that makes AI morally intelligent
D) A programming framework for ethical algorithms
Answer: B
Ethics = guidelines (not regulations/laws). It governs the creation and use of AI. It does NOT make AI morally intelligent (C) -- AI has no morals. It's not a programming framework (D).

UK AI Principles -- Deep Dive

🛡 Safety, Security & Robustness
AI systems must function safely, resist attacks, and perform reliably under varied conditions
🔍 Transparency & Explainability
Users should understand how and why an AI system made a particular decision
⚖ Fairness
AI should not discriminate and must treat individuals equitably regardless of background
📜 Accountability & Governance
Clear responsibility for AI systems; organizations must have governance structures in place
✍ Contestability & Redress
Individuals can challenge AI decisions and seek correction when harmed
🏢

AI Governance

AI Governance = A set of practices to keep AI systems safe, ethical, and under control

Includes:

  • Policies and standards
  • AI steering committees
  • Ethical review boards
  • Compliance frameworks
Exam Tip

Governance is about CONTROL and OVERSIGHT -- not technical implementation. Think management, not engineering.

Ethical Challenges to Behavior

💰 Self-interest
Personal gain influences decisions over ethical considerations
🔎 Self-review
Reviewing your own work without independent oversight
⚖ Conflict of Interest
Competing loyalties or financial interests compromise judgment
👎 Intimidation
Pressure or threats that prevent ethical behavior
📣 Advocacy
Promoting a position so strongly it compromises objectivity
Memory Trick

"SSCIA -- Some Selfish Conflicts Intimidate Advocates"

Strategies for Ethical AI

⚖ Dealing with Bias
Diverse teams, varied data sources, fairness metrics, regular audits
🔓 Openness
Open about methods, data, and limitations of AI systems
🔍 Transparency
Make AI decision-making processes visible and understandable
🤝 Trustworthiness
Build reliable, consistent AI systems that earn user confidence
💬 Explainability
AI decisions can be explained in human-understandable terms
Plus: Ethical Risk Framework

A framework that integrates ethical considerations from data collection through deployment

Next: SORT EXERCISE -- students sort items into challenges vs strategies
SORT EXERCISE

Challenge or Strategy?

Sort each item: is it an ethical CHALLENGE or an ethical STRATEGY?

Self-interest
Transparency
Conflict of interest
Explainability
Intimidation
Openness
Self-review
Trustworthiness
Advocacy
Dealing with Bias
Ethical Challenges (SSCIA)
  • Self-interest
  • Self-review
  • Conflict of interest
  • Intimidation
  • Advocacy
Ethical Strategies
  • Dealing with Bias
  • Openness
  • Transparency
  • Trustworthiness
  • Explainability

Exam Question

Which of the following poses a challenge to ethical behavior in AI?

A) Budgetary control processes
B) Transparency guidelines
C) Conflict of interest
D) Explainability standards
Answer: C
Conflict of interest is a CHALLENGE to ethical behavior. Transparency (B) and Explainability (D) are STRATEGIES for ethical behavior. Budgetary control (A) is not in the syllabus.

Exam Question

What is a strategy for addressing ethical challenges in AI?

A) Increasing automation speed
B) Reducing project documentation
C) Creating an ethical risk framework
D) Outsourcing AI development
Answer: C
An ethical risk framework integrates ethical considerations throughout the AI lifecycle. The other options either ignore ethics or are unrelated to ethical strategy.
📜

Why Regulate AI?

Regulation = manage associated risks and ensure accountability

Exam Trap

The purpose is NOT to limit innovation or slow down AI progress. It's to ensure clear legal accountability that governs effective management.

AI Regulation Landscape

UK GDPR
Data protection regulation with strict consent and processing rules
DPA 2018
UK Data Protection Act implementing GDPR domestically
WCAG
Web Content Accessibility Guidelines -- inclusive digital AI interfaces
EU AI Act (2024)
First comprehensive AI law -- risk-based classification system
ISO Standards
International standards for AI management, concepts, risk
NIST
US National Institute of Standards -- AI risk management framework
📋

ISO 42001 = AI Management System (AIMS)

ISO 42001 specifies requirements for an AI Management System

Classic Exam Trap

NOT ISO 9001 (quality), NOT ISO 22989 (AI concepts), NOT ISO 31000 (risk management). Only ISO 42001 = AIMS.

ISO 22989 = AI concepts & terminology

ISO 31000 = Risk management

ISO 9001 = Quality management

ISO 42001 = AI Management System

Exam Question

What ISO standard specifies requirements for an AI Management System (AIMS)?

A) ISO 9001
B) ISO 22989
C) ISO 31000
D) ISO 42001
Answer: D
ISO 42001 = AIMS. ISO 9001 = quality management. ISO 22989 = AI concepts and terminology. ISO 31000 = risk management. Remember: 42001 for AI Management.

Risk Management in AI

Risk
"A person or thing regarded as a threat or likely source of danger."

Risk is about what COULD go wrong -- threats, dangers, vulnerabilities.

Risk Management
"A series of processes which allow risk to be understood and minimized proactively."

Risk management is PROACTIVE -- identify and address risks BEFORE they occur.

Risk Management Techniques

📊 Risk Analysis
Identify, assess, and prioritize risks systematically
📈 SWOT
Strengths, Weaknesses, Opportunities, Threats
🌎 PESTLE
Political, Economic, Social, Technological, Legal, Environmental
🛠 Cynefin
Framework for decision-making in different complexity contexts
WARNING -- Exam Trap

"Crisis" is NOT a risk management technique! It's a situation. Risk Analysis, SWOT, PESTLE, and Cynefin ARE techniques.

Risk Mitigation Strategies

👥 Ownership & Accountability
Assign clear risk owners who are accountable for managing specific risks
🤝 Stakeholder Involvement
Engage stakeholders early and often in risk identification and response
🎓 Subject Matter Experts
Leverage domain expertise to identify and address specialized risks

Exam Question

Which is NOT a risk management technique?

A) Crisis
B) PESTLE
C) SWOT
D) Cynefin
Answer: A
Crisis is a SITUATION, not a technique. PESTLE, SWOT, and Cynefin are all legitimate risk management techniques from the EXIN syllabus. Risk Analysis is also a technique.
Topic 3
15%

Enablers of AI

Pace: FAST -- you know this

Common AI Examples

🤖 Human Compatible
AI designed to align with human values and preferences
⌚ Wearable
Smartwatches, fitness trackers with AI health monitoring
💻 Edge
AI processing locally on device, not in the cloud
📡 IoT
Sensors and connected devices feeding data to AI systems
💉 Personal Care
AI in healthcare monitoring and personal wellness
🚗 Self-Driving
Autonomous vehicles using AI for navigation and decision-making
✍ Generative AI
ChatGPT, DALL-E, Copilot -- creating content from prompts
🌎

AI You Already Use

📱 Face Unlock
Narrow AI image recognition on your phone
🎵 Spotify Discover
ML recommendation engine analyzing listening patterns
🌐 Google Translate
NLP converting text between 100+ languages
📧 Spam Filter
ML classification -- learning what's spam vs legitimate

Robotics in AI

Robotics Definition
"A machine that can carry out a complex series of tasks automatically, either with or without intelligence."

Types of Robots:

  • Industrial -- manufacturing, assembly lines
  • Personal -- household, companion
  • Autonomous -- self-directed operation
  • Nanobots -- microscale robots
  • Humanoids -- human-shaped robots
RPA (Robotic Process Automation)
A machine that can carry out a complex series of tasks automatically, either with or without intelligence, usually with a goal of improving processes.
Key Distinction

Robots can be intelligent OR non-intelligent. RPA automates repetitive tasks -- it may or may not involve AI.

Exam Question

What is a machine that can carry out complex tasks automatically?

A) Algorithm
B) Neural network
C) Database
D) Robotic
Answer: D
The definition of a robot is "a machine that can carry out a complex series of tasks automatically." An algorithm (A) is a set of instructions. A neural network (B) is a ML model. A database (C) stores data.

The AI Hierarchy

🤖 Artificial Intelligence
The broadest umbrella -- machines demonstrating intelligence
📈 Machine Learning
Subset of AI -- algorithms that improve through experience
🧠 Deep Learning
Subset of ML -- multi-layered neural networks
💬 Large Language Models (LLMs)
DL on massive datasets -- recognize, summarize, generate content

Key ML Definitions

Machine Learning
"The field concerned with the question of how to construct computer programs that automatically improve with experience." -- Tom Mitchell
Neural Network
"A program or model that makes decisions in a manner similar to the human brain, using processes that mimic the way biological neurons work together."
Deep Learning
"A multi-layered neural network." Multiple hidden layers enable learning complex patterns.
LLM
"Deep learning algorithms that can recognize, summarize, translate, predict, and generate content using very large datasets." -- IBM

Common ML Concepts

🎯 Prediction
Forecasting future outcomes based on historical data patterns
👁 Object Recognition
Identifying objects, faces, or patterns within images and video
📁 Classification
Categorizing data into predefined groups (includes random decision forests)
📐 Clustering
Grouping similar data points together without predefined categories
⭐ Recommendations
Suggesting relevant items based on user behavior (Netflix, Spotify)

Exam Question

Which of the following is a machine learning concept?

A) Bunching
B) Combining
C) Clustering
D) Clumping
Answer: C
Clustering is the correct ML term. Bunching, Combining, and Clumping are made-up distractors that sound similar but are not ML terminology.

Types of Machine Learning

Supervised

  • Labeled data
  • Known inputs AND outputs
  • Model learns correct answers
  • Example: email spam classification
|

Unsupervised

  • Unlabeled data
  • No predefined outputs
  • Model finds patterns
  • Example: customer segmentation
Semi-supervised Learning

Starts with a small amount of labeled data, then adds a larger amount of unlabeled data. Best of both worlds.

Next: CLASS POLL -- have students vote with fingers (1-4)
CLASS POLL

Vote Now!

A company trains its fraud detection model with 500 labeled transactions, then adds 200,000 unlabeled transactions. What type of learning is this?

A) Supervised Learning
B) Unsupervised Learning
C) Semi-supervised Learning
D) Reinforcement Learning
Answer: C) Semi-supervised Learning
Starts with small labeled dataset, then introduces larger unlabeled data. Textbook definition of semi-supervised learning.
🧠

Remember the Learning Types

🏫 Supervised
Teacher gives answers (labeled). Student checks work against known correct answers.
📚 Unsupervised
Student figures it out alone (unlabeled). No answer key -- find your own patterns.
🤝 Semi-supervised
Teacher starts with a few answers, student continues on their own with the rest.

Exam Question

What type of data does semi-supervised learning use?

A) Only labeled data
B) Labeled AND unlabeled data
C) Only unlabeled data
D) Pre-processed data only
Answer: B
Semi-supervised uses BOTH: a small amount of labeled data AND a larger amount of unlabeled data. Only labeled = supervised. Only unlabeled = unsupervised.
Topic 4
20%

Finding & Using Data in AI

Pace: Medium -- lists to memorize

Key Data Terms

Big Data
"Extremely large data sets that may be analyzed computationally to reveal patterns, trends, and associations." -- Dialogic.com
Data Visualization
"The representation of data through use of common graphics, such as charts, plots, infographics and even animations." -- IBM

Data Types

Structured

  • Organized in tables
  • Rows and columns
  • SQL databases
  • Spreadsheets

Sequential or serial tabular format

|

Unstructured

  • No predefined format
  • Images, video, audio
  • Social media posts
  • Free-form text

No pre-defined order or structure

Semi-structured

Not tabular, but HAS some organizational properties: JSON, XML, emails, HTML

⭐

5 Data Quality Characteristics

This is HEAVILY tested -- memorize all five

The Five Data Quality Characteristics

✅ Accuracy
Is the data correct? Free from errors and accurately represents reality.
📦 Completeness
Is all the data there? No missing values, fields, or records.
🔒 Uniqueness
Is it free from duplication? Each record appears only once.
🔄 Consistency
Is it free from conflict? No contradictions (e.g., text in a number field).
⏰ Timeliness
Is it current and available? Data is up-to-date when needed.
Next: MATCHING EXERCISE -- students match data problems to quality characteristics
MATCHING EXERCISE

Match the Problem to the Characteristic

Which data quality characteristic is violated in each scenario?

Duplicate customer records
Last year's data for current forecasts
Text in a number field
Customer name misspelled
30% of emails missing
Duplicate customer records
→
Uniqueness
Last year's data for current forecasts
→
Timeliness
Text in a number field
→
Consistency
Customer name misspelled
→
Accuracy
30% of emails missing
→
Completeness
🧠

Memorize: Data Quality

Accuracy Completeness Uniqueness Consistency Timeliness

"All Companies Use Consistent Tables"

Or just remember: ACUCT

Consequences of Poor Data Quality

❌ Errors & Inaccuracies
Wrong data in = wrong results out. AI models make incorrect predictions.
⚖ Bias
Biased training data creates biased AI systems that discriminate unfairly.
👎 Loss of Trust
Users lose confidence in AI systems that produce unreliable results.
💰 Financial Penalties
Regulatory fines, lost revenue, wasted investment from poor data decisions.

Exam Question

Which data quality check ensures data is free from conflict (such as text appearing in a number field)?

A) Accuracy
B) Consistency
C) Completeness
D) Uniqueness
Answer: B
Consistency = free from conflict. Text in a number field is a data conflict. Accuracy = correct values. Completeness = no missing data. Uniqueness = no duplicates.

Risks of Handling Data in AI

⚖ Bias
Mitigate with: multiple sources, diversity in teams, fairness metrics
📰 Misinformation
Mitigate with: check source reliability, subject matter expert (SME) review
⚙ Processing Restrictions
Organizational requirements, frameworks, and internal regulations
📜 Legal Restrictions
UK GDPR, DPA 2018 -- staying abreast of evolving legal requirements
🔬 Scientific Method
Using empirical methods to validate data and results

Exam Question

Which AI risk can subject matter experts (SMEs) help minimize?

A) Bias
B) Processing restrictions
C) Legal restrictions
D) Misinformation
Answer: D
SMEs verify accuracy and reliability of information -- they minimize MISINFORMATION. Bias is addressed through diverse teams and data sources, not specifically SMEs. Tricky but important distinction.

Uses of Big Data

🗃 Storage & Use
Collecting and organizing massive datasets for analysis
👥 Understanding the User
Insight into customer behavior, preferences, and purchasing patterns
⚙ Improving Process
Optimizing business operations through data-driven decisions
⭐ Improving Experience
Personalizing products and services based on user data
Real World

Netflix analyzes viewing patterns of 200M+ users to personalize what each subscriber sees -- that's big data driving improved experience.

Exam Question

What is a common use of big data in organizations?

A) Reducing server capacity
B) Eliminating all data processing needs
C) Gaining insight into user purchasing patterns
D) Replacing human decision-making entirely
Answer: C
Understanding user purchasing patterns is a key use of big data. Big data provides insight -- it doesn't replace humans (D), eliminate processing (B), or reduce servers (A).

Data Visualization Techniques

📝 Written
Reports, summaries, annotations -- text-based data presentation
🗣 Verbal
Presentations, briefings, narrated data stories
📊 Pictoral
Charts, graphs, plots, diagrams -- visual representations
🔈 Sounds
Sonification -- representing data through audio patterns
📈 Dashboards & Infographics
Multiple KPIs displayed simultaneously in real-time
👓 VR / AR
Immersive data exploration in virtual or augmented reality

Exam Question

Which visualization technique is best for displaying multiple KPIs at once?

A) Dashboard
B) Written report
C) Verbal presentation
D) Audio sonification
Answer: A
Dashboards display multiple KPIs simultaneously in real-time. Written reports (B) and presentations (C) are sequential. Sonification (D) is not suited for multiple simultaneous metrics.

Generative AI & LLMs

Generative AI
"Refers to deep-learning models that can generate high-quality text, images, and other content based on the data they were trained on." -- IBM

How it works:

  • Trained on huge volumes of data
  • Generates coherent, human-sounding language
  • NLP and image generation
Large Language Models (LLMs)
"Deep learning algorithms that can recognize, summarize, translate, predict, and generate content using very large datasets." -- IBM

Key mechanism:

  • Uses training to predict the next word in text
  • Chain of words forms coherent responses
  • Prompt engineering improves outputs
💬

Prompt Engineering

Prompt Engineering = designing specific instructions to get specific outputs from AI

👎 Weak Prompt
"Summarize this."
👍 Strong Prompt
"Summarize this in 3 bullet points for a CEO audience, focusing on financial impact."
Key Concept

More specific, detailed prompts produce more specific, robust responses. This is the core of prompt engineering.

Exam Question

What describes altering instructions to generate specific output from AI?

A) Data visualization
B) Neural network training
C) Prompt engineering
D) Algorithm optimization
Answer: C
Prompt engineering = designing/altering instructions (prompts) to get specific, desired outputs from generative AI. Data visualization (A) displays data. Neural network training (B) is model development. Algorithm optimization (D) is performance tuning.

The Machine Learning Process

1
Analyze the Problem
2
Data Selection
3
Data Pre-processing
4
Data Visualization
5
Select ML Model (Algorithm)
5a
Train the Model
5b
Test the Model
5c
Repeat (learn from experience)
6
Review
🧠

Remember the ML Process

Analyze Select data Pre-process Visualize Model (Train, Test, Repeat) Review

"Ana Selects Pre-Vis Models, Reviews"

A S P V M R -- Analyze, Select, Pre-process, Visualize, Model, Review

Next: ORDER EXERCISE -- class puts ML process stages in sequence
ORDER EXERCISE

Put the ML Process in Order

These stages are jumbled. What's the correct sequence?

Review
Data Selection
Train the Model
Analyze the Problem
Data Visualization
Test the Model
Data Pre-processing
Select Algorithm
Repeat
  1. Analyze the Problem
  2. Data Selection
  3. Data Pre-processing
  4. Data Visualization
  5. Select Algorithm
  6. Train the Model
  7. Test the Model
  8. Repeat
  9. Review

Exam Question

Which ML process stage represents data as images and graphs to gain insight?

A) Data selection
B) Data pre-processing
C) Data visualization
D) Model training
Answer: C
Data visualization is the stage where data is represented as images, graphs, and charts to gain insight. Selection (A) is choosing data. Pre-processing (B) is cleaning data. Training (D) is teaching the model.
Topic 5
20%

Using AI in Your Organization

Pace: SLOW -- most non-obvious content

AI Opportunities in Organizations

⚙ Automation
Rule-based, repetitive processes that follow clear logic -- perfect for AI
🔄 Repetitive Tasks
Reduce human input on routine work -- data entry, scheduling, reporting
✍ Content Creation
Generative AI for drafting, summarizing, translating, creating media
🚚

Good vs Bad AI Fit

👍 Good Fit
  • Route planning in logistics
  • Invoice processing
  • Scheduling based on contracts
  • Fraud detection patterns
👎 Bad Fit
  • Employee appraisals
  • Emotional support
  • Creative strategy
  • Complex ethical judgment

Exam Question

Which task best represents process automation in a logistics company?

A) Conducting employee performance reviews
B) Creating rotas based on contracts and recurring deliveries
C) Developing company culture initiatives
D) Negotiating supplier contracts
Answer: B
Creating rotas from contracts and recurring deliveries is rule-based, repetitive, and data-driven -- perfect for AI automation. The other tasks require human judgment, negotiation, or creative thinking.
💼

The Business Case

A business case provides justification for undertaking a project and is used to secure funding

Purpose

Provides insight and justification. Gives decision makers enough detail to evaluate the proposed recommendations.

Business Case Structure

1. Introduction
Context, purpose, scope of the proposed project
2. Management Summary
Overarching brief for executives -- the big picture
3. Current State
Description of the as-is situation and why change is needed
4. Options Considered
Option described, cost/benefit analysis, impact assessment, risk assessment
5. Recommendations
Proposed solution with rationale and justification
6. Appendices
Supporting information, data, references
Next: BUILD IT -- class names the 6 business case sections one by one
BUILD IT TOGETHER

Name the 6 Business Case Sections

Call them out! Each click reveals the next section.

1Introduction
2Management / Executive Summary
3Description of Current State
4Options Considered
5Recommendations
6Appendices / Supporting Information

Business Case: Options Considered

📄 Option Described
Clear description of each alternative being evaluated
💰 Cost / Benefit Analysis
Financial comparison of costs versus expected benefits
📈 Impact Assessment
Effects on organization, processes, people, and technology
⚠ Risk Assessment
Identification and evaluation of potential risks for each option
🧠

Business Case Structure

Introduction Management Summary Current State Options Considered Recommendations Appendices

"IMCORA -- It Makes Current Options Really Accessible"

Exam Question

In a business case, where should an overarching brief for executives be documented?

A) Introduction
B) Management Summary
C) Current State
D) Appendices
Answer: B
The Management (or Executive) Summary is where the overarching brief goes. It provides executives with the big picture. The Introduction (A) sets context. Current State (C) describes the as-is. Appendices (D) are supporting documents.
👥

Stakeholders

Stakeholder = anyone who has influence on OR is impacted by the project

Key Point

Stakeholders include BOTH those who influence AND those who are affected. It's a two-way relationship. Tools for categorization: Power/Interest Grid and Stakeholder Wheel.

Stakeholder Power/Interest Grid

High Power / Low Interest
Keep Satisfied
High Power / High Interest
Constant Active Management
Low Power / Low Interest
Monitor
Low Power / High Interest
Keep Informed
Exam Favorite

High Power + High Interest = Constant Active Management. This exact phrase is tested.

Next: BUILD IT -- class fills in the Power/Interest Grid quadrants
BUILD IT TOGETHER

Fill the Power/Interest Grid

What goes in each quadrant? Call it out!

Low Power + Low InterestMonitor
Low Power + High InterestKeep Informed
High Power + Low InterestKeep Satisfied
High Power + High InterestConstant Active Management

Exam Question

How should a stakeholder with high power and high interest be managed?

A) Constant Active Management
B) Keep Satisfied
C) Keep Informed
D) Monitor
Answer: A
High power + high interest = Constant Active Management. Keep Satisfied (B) = high power, low interest. Keep Informed (C) = low power, high interest. Monitor (D) = low power, low interest.

Project Management Approaches

Agile

  • Iterative & incremental
  • Flexible, adapts to change
  • Short sprints
  • Continuous feedback
  • Best for: evolving requirements
|

Waterfall

  • Sequential, linear stages
  • Rigid, planned upfront
  • Each phase completes before next
  • Detailed documentation
  • Best for: fixed, well-defined projects
Hybrid

Combines Agile flexibility with Waterfall structure. Uses both approaches where each fits best.

Waterfall Stages (In Order)

1
Requirements
2
Design
3
IMPLEMENTATION
4
Testing
5
Deployment
6
Maintenance
Exam Trap

After Design comes Implementation (NOT Testing!). This is a frequently tested sequence question.

Next: SPOT THE MISTAKES -- find the errors in this Waterfall sequence
SPOT THE MISTAKES

Find the Errors in This Waterfall

This Waterfall sequence has mistakes. Can the class find them?

Requirements → Design → Testing → Implementation → Maintenance → Deployment

Correct: Requirements → Design → Implementation → Testing → Deployment → Maintenance

Two swaps: Testing/Implementation were swapped, and Maintenance/Deployment were swapped. Remember: "Robert Designed It, Then Deployed Monday"

🧠

Waterfall Stages Mnemonic

Requirements Design Implementation Testing Deployment Maintenance

"Robert Designed It, Then Deployed Monday"

Exam Question

In the Waterfall methodology, what phase comes after Design?

A) Testing
B) Implementation
C) Deployment
D) Requirements
Answer: B
The sequence is Requirements > Design > Implementation > Testing > Deployment > Maintenance. Testing (A) comes AFTER Implementation, not directly after Design.

Risk in AI Projects

Risk Assessment
The process of identifying, analyzing, and evaluating risks associated with an AI project

Identify risks BEFORE they become problems

Risk Owner
The INDIVIDUAL who is ultimately accountable for managing a specific risk
Exam Tip

Risk owner = the individual (not team, not department) who is ultimately accountable.

Risk Appetite & Management Strategies

Risk Appetite

The level of risk an organization is willing to accept in pursuit of its objectives

✅ Accept
Acknowledge the risk and live with it -- impact is acceptable
🛡 Mitigate
Reduce likelihood or impact through sharing, contingency planning
🚫 Avoid
Eliminate the risk entirely by changing plans or approach
🔁 Transfer
Shift risk to a third party -- insurance, outsourcing
🧠

Risk Strategies: AMAT

Accept Mitigate Avoid Transfer

"Always Mitigate And Transfer"

4 strategies. Accept, Mitigate, Avoid, Transfer. That's the complete list.

Financial Analysis & Triple Bottom Line

Financial Concepts

Forecasting
Predicting future costs and revenues
Margin for Error
Buffer for unexpected costs or shortfalls
Cost-Benefit Analysis
Systematic comparison of costs vs expected rewards

Triple Bottom Line (3Ps)

Profit + People + Planet
A framework for assessing the full impact of AI initiatives across three dimensions: financial returns, social impact, and environmental sustainability.
Exam Definition

"The framework for assessing impact of the 3Ps: Profit, People, Planet"

Next: CLASS POLL -- scenario-based risk strategy question
CLASS POLL

Vote Now!

A healthcare company outsources its AI model hosting to a cloud provider, transferring the operational risk and liability. Which risk management strategy is this?

A) Accept
B) Mitigate
C) Avoid
D) Transfer
Answer: D) Transfer
Outsourcing to a third party transfers the risk. Accept = live with it. Mitigate = reduce likelihood/impact. Avoid = eliminate the activity entirely.

Exam Question

Which framework assesses impact across Profit, People, and Planet?

A) SWOT Analysis
B) Cost-Benefit Analysis
C) Triple Bottom Line
D) PESTLE Analysis
Answer: C
Triple Bottom Line = Profit + People + Planet (the 3Ps). SWOT (A) is Strengths/Weaknesses/Opportunities/Threats. Cost-Benefit (B) is financial comparison. PESTLE (D) is macro-environmental analysis.

Exam Question

What describes the role of a risk owner?

A) The individual ultimately accountable for managing a specific risk
B) The team responsible for all project risks
C) The department that funds risk mitigation
D) The external auditor who reviews risk reports
Answer: A
Risk owner = the INDIVIDUAL (not team or department) who is ULTIMATELY accountable (not just responsible) for managing a specific risk.

AI Governance Activities

📜 Compliance
Satisfy regulations -- ensuring AI systems meet legal and regulatory requirements
🛡 Risk Management
Proactively detect and mitigate risks throughout the AI lifecycle
🔄 Lifecycle Governance
Manage, monitor, and govern AI models from development to retirement
🧠

Governance: CRL

Compliance -- satisfy regulations Risk Management -- detect & mitigate Lifecycle Governance -- manage, monitor, govern

Three governance pillars. Nothing more, nothing less.

Exam Question

Which governance area identifies and reduces threats to AI systems?

A) Compliance
B) Lifecycle governance
C) Competitor analysis
D) Risk assessment
Answer: D
Risk assessment identifies and reduces threats. Compliance (A) satisfies regulations. Lifecycle governance (B) manages models over time. Competitor analysis (C) is not a governance activity at all.

Exam Question

Which is a governance activity required when implementing AI?

A) Competitor analysis
B) Cost-benefit analysis
C) Market research
D) Risk assessment
Answer: D
Risk assessment is a governance activity (under Risk Management). Competitor analysis (A), cost-benefit analysis (B), and market research (C) are business analysis activities, not governance activities.
Topic 6
15%

Future Planning & Impact

Pace: Normal

AI-Specific Career Roles

💻 ML Engineer
Builds and deploys machine learning models
📊 Data Scientist
Extracts insights from data using statistical methods
🔬 AI Research Scientist
Advances fundamental AI knowledge and algorithms
👁 Computer Vision Engineer
Develops systems that interpret visual information
💬 NLP Engineer
Builds systems that understand and generate human language
🤖 Robotics Engineer
Designs and builds intelligent robotic systems
⚖ AI Ethics Specialist
Ensures AI is developed and used responsibly
🌎 AI Anthropologist
Studies human-AI interaction and societal impact
💼

AI Transforms Existing Roles Too

AI doesn't just create new roles -- it transforms existing ones

  • Additional training -- new skills and knowledge
  • Improved efficiency -- AI-augmented workflows
  • Automation -- routine tasks handled by AI
  • Role evolution -- shifting from execution to oversight

AI Applications by Sector

📈 Marketing
Automating data analysis, trend prediction, targeted campaigns
🏥 Healthcare
Medical diagnostics, drug discovery, patient monitoring
🏦 Finance
Fraud detection, algorithmic trading, loan evaluation
🚗 Transportation
Self-driving vehicles, route optimization, traffic management
🎓 Education
Personalized learning, automated grading, adaptive content
🏭 Manufacturing
Quality control, predictive maintenance, supply chain optimization
🎬 Entertainment
Recommendation algorithms, content generation, personalization
💻 IT
AI-powered chatbots, digital assistants, automated testing

Exam Question

How does AI enhance marketing?

A) Automating data analysis and trend prediction
B) Replacing all marketing staff
C) Eliminating the need for market research
D) Guaranteeing sales outcomes
Answer: A
AI enhances marketing by automating analysis and predicting trends. It doesn't replace staff (B), eliminate research (C), or guarantee outcomes (D). AI augments, not replaces.

Exam Question

Which generative AI tool is commonly used in sales and marketing?

A) Chatbots
B) Industrial robots
C) Self-driving vehicles
D) Medical diagnostic systems
Answer: A
Chatbots are generative AI tools used in sales and marketing for customer interaction, lead qualification, and support. The other options are AI applications in different sectors entirely.

AI: Benefits vs Challenges

👍 Benefits

  • Reduces human error
  • Processes vast data for insights
  • Assists in medical diagnosis
  • Increases efficiency and productivity
  • 24/7 availability
vs

👎 Challenges

  • Algorithm bias and privacy
  • Job displacement
  • No empathy or creativity
  • Security risks (hacking)
  • Socio-economic inequality
  • Market volatility from AI trading

Environmental & Economic Impact

🌱 Environmental

  • Energy consumption of data centers
  • Water usage for cooling
  • Carbon footprint of training models
  • Need for sustainable AI practices
  • Climate change implications

💰 Economic

  • Job displacement in some sectors
  • Creation of new AI roles
  • Need for retraining and upskilling
  • Market disruption and volatility
  • Wealth concentration concerns
🤝

Human Plus Machine

Human + Machine = augmenting capabilities together

Critical Distinction

NOT competition. NOT replacement. It's COMBINATION to augment and enhance human capabilities with AI.

Exam Question

'Human plus machine' refers to what?

A) AI replacing human workers entirely
B) Combination of human and AI to augment capabilities
C) Humans competing against AI systems
D) Teaching machines to become human-like
Answer: B
'Human plus machine' = combination to augment capabilities. Not replacement (A), not competition (C), not making machines human (D). The key word is AUGMENT.

Consciousness in AI

Human Consciousness
Sentience -- subjective experience, self-awareness, the ability to feel and experience the world. We know we exist.
AI Consciousness
Hypothetical -- can AI have autonomous intentions? Can it make conscious decisions? Can it have subjective experience?
Key Distinction

Functional capabilities (mimicking consciousness) vs genuine consciousness (actually experiencing).

"A future period characterized by rapid technological growth that will irreversibly transform human life."
-- Kurzweil, describing the Singularity
Kurzweil's Singularity
A point where AI surpasses human intelligence, leading to exponential, irreversible change.
Seth's Theory
Predictive processing, perception, and the nature of self -- consciousness emerges from the brain's predictions about the world.

Exam Question

What term describes rapid technological change that will irreversibly transform human life?

A) AI Winter
B) Digital Transformation
C) Industrial Revolution
D) Singularity
Answer: D
Singularity = rapid technological growth that irreversibly transforms human life (Kurzweil). AI Winter (A) = period of reduced AI funding/interest. Digital Transformation (B) and Industrial Revolution (C) are broader concepts.
Review

Final Review & Exam Prep

Rapid fire -- test yourself

Top Exam Traps (1-5)

⚠ Trap 1
Wrong: "Domain AI" Right: Narrow AI / Weak AI / ANI
⚠ Trap 2
Wrong: ISO 9001 for AIMS Right: ISO 42001 = AI Management System
⚠ Trap 3
Wrong: Crisis is a technique Right: Crisis is a situation, not a risk technique
⚠ Trap 4
Wrong: Testing after Design Right: Implementation comes after Design in Waterfall
⚠ Trap 5
Wrong: Regulation limits innovation Right: Regulation manages risk and ensures accountability

Top Exam Traps (6-10)

⚠ Trap 6
Wrong: Ethics = law Right: Ethics are voluntary guidelines; laws are mandatory
⚠ Trap 7
Wrong: SMEs fix bias Right: SMEs minimize misinformation. Diversity fixes bias.
⚠ Trap 8
Wrong: ChatGPT = AGI Right: All current AI (including ChatGPT) is Narrow AI
⚠ Trap 9
Wrong: Human+Machine = replacement Right: Human+Machine = augmenting capabilities together
⚠ Trap 10
Wrong: Challenges = Strategies Right: Challenges (SSCIA) ≠ Strategies (bias, openness, transparency...)

6 Most-Tested Definitions

Scientific Method
An EMPIRICAL method for acquiring knowledge
Risk
A threat or likely source of danger
AI Governance
Practices to keep AI safe, ethical, and under control
Stakeholder
Anyone who influences or is impacted by the project
Ethics
Moral principles governing behavior or activity
Prompt Engineering
Designing specific instructions for specific AI outputs

All Memory Tricks

Floridi & Cowls
BNAJE: "Be Nice And Just Explain"
UK AI Principles
STFAC: Safety, Transparency, Fairness, Accountability, Contestability
Data Quality
ACUCT: "All Companies Use Consistent Tables"
Ethical Challenges
SSCIA: "Some Selfish Conflicts Intimidate Advocates"
Business Case
IMCORA: "It Makes Current Options Really Accessible"
Waterfall
RDITDM: "Robert Designed It, Then Deployed Monday"
ML Process
ASPVMR: "Ana Selects Pre-Vis Models, Reviews"
Risk Strategies
AMAT: Accept, Mitigate, Avoid, Transfer
🎯

Your Exam Strategy

Before the Exam

  • Take practice exams under timed conditions
  • Focus extra time on Topics 2 & 5
  • Memorize the cheat sheet
  • Review all mnemonics
  • Know exact definitions

During the Exam

  • Read ALL four options before answering
  • Watch for absolute words ("always", "never", "all")
  • Eliminate obviously wrong answers first
  • Flag uncertain questions and return later
  • 90 seconds per question -- don't linger

Rapid Fire -- 5 Questions

Test yourself. Click to reveal each answer.

Q1: What ISO standard = AIMS? ISO 42001

Q2: High power + high interest stakeholder strategy? Constant Active Management

Q3: What comes after Design in Waterfall? Implementation

Q4: 5 data quality characteristics? Accuracy, Completeness, Uniqueness, Consistency, Timeliness

Q5: Human + Machine = ? Augmenting capabilities together

🏆

Go ace that exam!

You've covered all 6 topics, 40+ key concepts, and 20+ exam-style questions.

40 Questions | 60 Minutes | 65% to Pass

Remember: ACUCT, IMCORA, AMAT, CRL

Good luck. You've got this.