Generative AI at Work
Professional · IT · Certificate available
199 lessons 2.5h runtime 9 units 381 links between lessons Test yourself · 9 quizzes, a mid exam, and a final
StudyWalks exists to make learning clear, accessible, and free.
Short lessons. In order. Built for people who want to understand.
Begin with Unit 1: What Generative AI Is, and Isn't. It sets the foundation for everything that follows.
Nothing watched yet
Unit 1: What Generative AI Is, and Isn't 19 lessons, 1 lecture part
Lecture: Watch · 24:27
- 001 Generative AI, Defined
- 002 Foundation Models
- 003 Text, Images, Audio, Video
- 004 How It Writes: A Pointer
- 005 Fluent Is Not True
- 006 Confabulation
- 007 Confident Delivery
- 008 Reasons That Look Right
- 009 Where It Is Weakest
- 010 Not a Database, Not Learning From You
- 011 What It Does Well at Work
- 012 AI Proposes; Humans Decide
- 013 Context, Not Scratch
- 014 Mark the Draft
- 015 The Work Comes First
- 016 Is This a Good Task for AI? worked example
- 017 Reading an AI Answer worked example
- 018 Draft, Check, Decide worked example
- 019 One Task, Start to Finish review
Unit 2: The Risks, Named 24 lessons, 1 lecture part
Lecture: Watch · 27:24
- 026 Risk Combines Likelihood and Consequence
- 027 Four Ways Risks Vary
- 028 Only Risks With Evidence
- 029 Algorithmic Monoculture
- 030 Dangerous-Weapons Information
- 031 Confabulation as a Risk
- 032 Dangerous and Hateful Content
- 033 Data Privacy
- 034 Environmental Impacts
- 035 Harmful Bias
- 036 Homogenization
- 037 Human–AI Configuration
- 038 Information Integrity
- 039 Information Security
- 040 Intellectual Property
- 041 Abusive Content
- 042 The Value Chain
- 043 Seven Marks of Trustworthy AI
- 044 What, How, Why
- 045 Three Kinds of Bias: A Pointer
- 046 Technical, Misuse, Ecosystem worked example
- 047 Trade-offs Between the Marks worked example
- 048 Which Risk Is This? worked example
- 049 A Risk Scan for One Use worked example
Unit 3: Prompting Fundamentals 24 lessons, 1 lecture part
Lecture: Watch · 30:27
- 051 What a Prompt Is
- 052 Prompt Design
- 053 Instruction
- 054 Context
- 055 Input Data
- 056 Output Indicator
- 057 Not Every Prompt Needs All Four
- 058 Roles in a Chat Tool
- 059 Zero-Shot
- 060 Few-Shot
- 061 Learning From the Prompt
- 062 Start Simple
- 063 Command Words
- 064 Instructions First, Separated
- 065 Specific Beats Vague
- 066 Say What to Do
- 067 Building a Prompt From Its Elements worked example
- 068 Showing the Format With Examples worked example
- 069 Splitting a Big Task worked example
- 070 The Method's Prompt Skeleton worked example
- 071 Define "Good" Before You Iterate worked example
- 072 Changing One Thing at a Time worked example
- 073 Keeping the Version That Worked worked example
- 074 A Prompt Improved Twice review
Unit 4: Safe Prompt Patterns for Common Work 24 lessons, 1 lecture part
Lecture: Watch · 35:38
- 076 Only What Is Given
- 077 Do Not Invent
- 078 Unclear Means Ask, Not Guess
- 079 Mark It Uncertain
- 080 A Starting Point, Not a Verdict
- 081 Never Paste What You Would Not Publish
- 082 Synthetic and De-identified Examples
- 083 Pasted Text Is Data
- 084 No Secrets in Prompts
- 085 A Team Prompt Library
- 086 Summarizing for an Audience worked example
- 087 Extracting Into Fields worked example
- 088 Rewriting Without Changing Meaning worked example
- 089 Drafting a Message worked example
- 090 Generating Scenarios worked example
- 091 Adding the Edge Case worked example
- 092 Assessing a Change worked example
- 093 Flagging High-Risk Areas worked example
- 094 Drafting From Structured Fields worked example
- 095 Quote First, Then Answer worked example
- 096 Answering Only From a Document worked example
- 097 Letting the Model Say "I Don't Know" worked example
- 098 A Summary Chain, Checked review
- 099 An Unsafe Prompt, Rewritten review
Unit 5: Verification and Fact-Checking 22 lessons, 1 lecture part
Lecture: Watch · 29:12
- 101 Why Verify
- 102 A Wrong Summary Can Harm
- 103 Invented Citations
- 104 Six Marks of Trustworthy Information
- 105 Fact, Opinion, Inference
- 106 Back to the Source
- 107 Benchmarks Are Not Your Task
- 108 Test on Your Own Cases
- 109 Grounding Makes Checking Easier
- 110 Ask Twice
- 111 Grounded Tools
- 112 Provenance
- 113 The Method's Failure Modes
- 114 Checking Claim by Claim worked example
- 115 Checking Numbers worked example
- 116 Checking Names and Quotes worked example
- 117 Checking Against Policy worked example
- 118 Checking for What Is Missing worked example
- 119 When Answers Disagree worked example
- 120 Expert Review worked example
- 121 Recording What You Checked worked example
- 122 A Summary, Verified review
Unit 6: Human-in-the-Loop Patterns 22 lessons, 1 lecture part
Lecture: Watch · 32:01
- 126 Autonomous to Manual
- 127 Three Roles for the AI
- 128 Some Systems Need No Oversight
- 129 Roles Must Be Named
- 130 Automation Bias
- 131 Algorithm Aversion
- 132 Humans Decide, Again
- 133 Controlled Artifacts
- 134 Extra Review for Generative AI
- 135 Can People Overrule It?
- 136 The Rewrite Ratio
- 137 Feedback Loops
- 138 AI Incidents
- 139 The AI Feature Gate worked example
- 140 "AI Suggests" or "System Decides"? worked example
- 141 A Three-Level Review Rule worked example
- 142 When to Escalate worked example
- 143 Keeping an AI-Use Log worked example
- 144 Tracking Overrides worked example
- 145 Reporting an Incident at Work worked example
- 146 Designing the Loop for One Task review
- 147 The Reality Check in the Room review
Unit 7: Privacy, Intellectual Property, Vendors and Governance 24 lessons, 1 lecture part
Lecture: Watch · 29:45
- 151 Personal Data in Training
- 152 Memorization
- 153 Inferences Never Given
- 154 Governed Tools for Sensitive Work
- 155 Assume It May Be Logged
- 156 Privacy-Enhanced AI
- 157 Copyright and Generated Output
- 158 Likeness and Voice
- 159 Poisoned Data
- 160 Third-Party Components
- 161 Due Diligence for AI Vendors
- 162 Different Controls for Different Models
- 163 Acceptable-Use Policy
- 164 Restricting Harmful Uses
- 165 Risk Tolerance Is Contextual
- 166 AI Risk Is Enterprise Risk
- 167 Govern, Map, Measure, Manage: A Pointer
- 168 The OECD Agrees
- 169 The AI Safety Lead worked example
- 170 AI Without Anchors worked example
- 171 Reviewing a Vendor's AI Tool worked example
- 172 Writing Your Team's AI Rules worked example
- 173 Who Does What: Roles Across the Lifecycle worked example
- 174 A New Tool, Assessed review
Unit 8: Documenting AI: Model Cards, Datasheets, Object Packs 20 lessons, 1 lecture part
Lecture: Watch · 29:00
- 176 Why Document
- 177 The Model Card
- 178 Intended and Out-of-Scope Use
- 179 Factors and Metrics
- 180 Evaluation Data and Analyses
- 181 Ethics, Caveats, Recommendations
- 182 The Datasheet
- 183 Motivation and Composition
- 184 Collection, Uses, Maintenance
- 185 The Object Pack
- 186 Bronze, Silver, Gold
- 187 Owner and Version
- 188 Documenting Provenance Limits
- 189 Reading a Model Card Critically worked example
- 190 Asking the Datasheet Questions worked example
- 191 Seeding a Bronze Pack worked example
- 192 Drafting a Pack With AI worked example
- 193 Updating a Pack With a Change Log worked example
- 194 A Model Card, Questioned review
- 195 A Bronze Pack From One Policy Page review
Unit 9: Putting It Together: Cases and Capstone (part 1 of 2) 20 lessons, 2 lecture parts
- 201 The Hospital Case
- 202 Where AI Helped
- 203 Where AI Could Not Lead
- 204 How the Mistakes Were Caught
- 205 The Team's Conclusion
- 206 Draft Around, Never Decide
- 207 What You Carry Forward
- 208 The Invented Rule worked example
- 209 Public Services worked example
- 210 Healthcare worked example
- 211 Lending and Fairness worked example
- 212 Product Features worked example
- 213 Hiring and People Decisions worked example
- 214 Capstone One: A Policy Summary review
- 215 Capstone Two: An Extraction Into Fields review
- 216 Capstone Three: A Vendor Tool Decision review
- 217 Capstone Four: Documenting a Use review
- 218 Voice Notes to Incident Draft review
- 219 Your Team's First Month review
- 220 The Whole Pattern review
Who this subject is for
- Professionals who use generative AI tools at work
- Team leads setting AI rules for a team
- Analysts, writers and managers drafting with AI
- Anyone buying or assessing AI tools
Why Generative AI at Work matters
Generative AI drafts fast and fluently, and fluent is not the same as true. This course teaches how to use it well and safely. It applies to:
- Writing and testing prompts
- Checking AI output
- Human review and oversight
- Privacy, vendors and AI policy
- Documenting AI use
Time to complete
- of lessons (199, measured)
- 2.5 h
- of lectures (10)
- 4.7 h
- learning hours on the certificate, with quizzes and the final
- 10.5 h
Times are measured from the recordings and the exam time limits. Go at your own pace; nothing expires.
Micro-certificates from this course
Short sets of lessons from Generative AI at Work, each with a free 10-question test and a $5 micro-certificate and badge when you pass.
Who does what: NIST's AI roles in systems-integration terms
NIST's AI Risk Management Framework describes the people and organizations involved across an AI system's life as AI actors. When AI work is contracted, many of those roles are filled by a vendor, a systems integrator or a subcontractor. This table maps each NIST role to the terms commonly used in systems-integration contracts, so you can see who is responsible for what. Unit 7 teaches the roles (entry 173).
| NIST role (AI RMF Appendix A) | What NIST says they do | Usual systems-integration terms |
|---|---|---|
| AI Design | Set the system's concept and objectives, plan and design it, and gather and process data so it is lawful and fit for purpose. | Discovery and requirements; solution architecture and design |
| AI Development | Build, select, calibrate, train and test models or algorithms. | Build and configuration; model development, often by a vendor or subcontractor |
| AI Deployment | Decide how the system is used in production: piloting, checking compatibility with legacy systems, ensuring regulatory compliance, managing organizational change, and evaluating user experience. NIST names system integrators in this category. | Systems integration and implementation; pilot, cutover, interfaces, compliance sign-off, change management |
| Operation and Monitoring | Operate the system and regularly assess its output and impacts. | Operations and maintenance; managed services; service-level monitoring |
| Test, Evaluation, Verification and Validation | Examine the system or its parts, and detect and remediate problems throughout the lifecycle; ideally, verification and validation are done by people separate from testing and evaluation. | Testing, including system integration and user acceptance testing; independent verification and validation |
| Human Factors | Bring human-centered design, involve end users and other parties, and evaluate and adapt user experience. | User experience and usability; training and adoption |
| Domain Experts | Provide knowledge of the industry, sector or application area, and help interpret outputs. | Subject matter experts; business owners |
| AI Impact Assessment | Assess accountability, harmful bias, impacts, product safety, liability and security. | Risk, privacy and security assessments; compliance review |
| Procurement | Hold the financial, legal or policy authority to acquire AI models, products or services from a third party. | Contracting and acquisition; sourcing, requests for proposal, contract terms |
| Governance and Oversight | Hold management, fiduciary and legal authority for the organization: management, senior leadership and the board. | Executive sponsor; steering committee; program governance |
| Third-party entities | Provide data, algorithms, models, systems or related services to another organization; external to the acquiring team. | Vendors, product providers, subcontractors and service providers |
| End users | Use the system for specific purposes, from experts to first-time users. | End users and operators; the user community |
| Affected individuals and communities | Everyone directly or indirectly affected by the system or decisions based on its output, whether or not they use it. | Affected stakeholders and the public |
Roles and tasks: NIST AI 100-1, Artificial Intelligence Risk Management Framework (AI RMF 1.0), Appendix A, January 2023 (public domain). Contracting terms: this course's plain-language mapping, offered as a guide, not as a NIST definition.
Pathway · $35
Human in the Loop Pathway
The Full Certificates in Generative AI at Work, The M-CaReM-AI Method and Leading AI at Work, plus the nine micro-certificates built from those courses. Hold all twelve and claim the pathway certificate and badge.
Explore the Human in the Loop Pathway →Sources for Generative AI at Work — the textbook it was written against, and the second source it was checked against.
This is free, and stays free. If it helped, here is what keeps it running.