About 4 hours · Human-marked final assessment
AI at Work Foundations
Everything you should know before you use AI tools at work: the rules you work inside, what these tools are and where they fail, what they can already reach, and what stays your job. Written for New Zealand workplaces, with a final assessment marked by a person.
Course Overview
Section 1
Where the work actually goes
What you will be able to do: See where your working day actually goes, and why making one step faster changes little unless it is the step where the work waits.
The day that disappears
Where a working day actually goes: routine work, rework and the waits between people. Why a full day and a productive one are not the same thing.
Knowledge stuck in heads
What happens when the one person who knows is away, and why six versions of the same form are the same problem. Expertise that is never written down belongs to a person, not the business.
Disconnected systems and the copy and paste tax
The same facts retyped into four places, and the cost nobody budgets for. Why making one person faster changes little when the work is waiting at a handoff.
Faster is not the same as better
Adoption speeds up the work you already do; adaptation changes it. And the honest half: the time a tool saves in drafting, the checking can take back.
1 quiz
Section 2
The rules you work inside
What you will be able to do: Know the rules you work inside before you are given a tool: what you may put into one, what it can already reach, where your obligations come from, and who to ask when you are not sure. Inside those lines you can move without asking.
What AI guidelines and policies are for
Why organisations have AI guidelines, and why they give permission as much as they set limits. Where to find yours, and what to do if there are none.
The Four Gates
Four questions to ask, in order, before you paste or upload anything into an AI tool. The first no or not sure stops you, and you learn what to do next.
Commercial risk is not the same as personal information
Two kinds of sensitive information, two different consequences and two different people to ask. Plus the mistake that catches people out: material you paid for but do not own.
Three things that stay true whatever tool you use
The Privacy Act applies unchanged, confidentiality is a contract rather than a setting, and accountability does not delegate, whatever tool you use.
Whose permissions is it using?
What an AI tool can already reach without you pasting anything, and why that can be wider than you think. You cannot tell by looking, so you learn what to ask.
When a gate stops you
The escalation route for the whole course: who you ask, how fast they answer and what happens if nobody does. And what to do when something gets through that should not have.
Where the rules actually come from
Five sources of rules besides your AI guidelines, from your employment agreement to the Privacy Act. Knowing which to check, and when it is a question for someone else.
1 quiz · 1 practical exercise
Section 3
What you are actually working with
What you will be able to do: Explain what an AI tool is and how it fails, using three images you can hold, and why your knowledge of your own work stays the final check.
A person, a puppet, a shadow
Three ways to see an AI tool, drawing on Dr Karaitiana Taiuru's Kaupapa Māori AI Framework. Why a tool that talks like a person gets trusted like one, and why that is where most mistakes start.
Whose hand is on the strings
Why the same AI model answers differently depending on who set it up, and the forces that shape every output. Many hands are involved; the organisation and the person who used it still own the result.
A shadow has no light of its own
An AI tool is made from what people have written, so it is weakest where little has been written: New Zealand specifics, recent changes, your own industry. And it sounds just as sure.
Where it fails, and why it agrees with you
Confident errors that look exactly like right answers, and a tool that leans towards agreeing with you. How to ask in ways that bring the gaps to the surface.
You are the subject matter expert
What memory and saved context change, and what no tool can have: your knowledge of your own work. Why that knowledge is the final check.
1 quiz · 1 practical exercise
Section 4
What the tools can reach
What you will be able to do: Know the kinds of AI you will meet, what the main tools can already reach, who owns what they produce, and which tool and account to use for work.
The kinds of AI you will meet
Language, speech, image, vision and music tools, with everyday examples. Most workplace AI reads and writes text, but it is not all chatbots.
The four you will actually use, and what they can already see
ChatGPT, Copilot, Claude and Gemini: what differs is where your data goes and what each can reach. Why the account matters more than the brand.
Who owns the output
Who owns what you generate, what watermarks and provenance markers tell you, and what happens to what you upload, including what New Zealand law has not settled yet.
The notetaker in the room
Meeting notetakers are often the AI already in the room, and the least discussed. Who connected it, who is recorded, where the transcript goes and who can read it later.
Which tool, and which account
The one rule in the course about accounts: approved tools, or new ones that have been through approval, and never a personal account for the organisation's work or data. The five reasons why, and how to ask for a tool.
1 quiz
Section 5
What you put in, and what you sign off
What you will be able to do: Decide what to hand to a tool and what to keep, check what comes back against the source, and name who is accountable before anything is sent.
The Delegation Grid
Two questions that decide what to hand to a tool: can you undo it, and can you check it? If you can do neither, do it yourself.
Verification at source
Why reading a paragraph is not checking it, and five questions that find out whether an output holds up.
The Three Checkpoints
Input, judgement and commitment, and why most workflows staff only the last one. A signature is not a review.
It was not the AI that sent it
Whatever drafted the work, the person who sends it answers for it. Plus the AI Use Checklist, which pulls the rules so far onto one card.
1 quiz · 1 practical exercise
Section 6
When AI does things, not just drafts them
What you will be able to do: Tell a tool that drafts for you from one that acts on its own, and know the one rule and four questions that keep any of them safe near real records.
Ask and act, or brief and done
The difference between a tool that waits for you and one that carries out an objective on its own, and where the risk changes. If you cannot explain a result you carry, you cannot use it.
Prompts, skills, agents and plain old automation
Four words people use interchangeably, and the line between them. The one rule that keeps any of them safe near real records, and four questions to ask before anyone builds an agent.
1 quiz
Section 7
The Org Brain
What you will be able to do: See why a tool's answers depend on what your organisation has written down, and ask of any task what it would need to know.
Four layers of knowledge
From a single prompt to knowledge the whole organisation can draw on. Why a tool knows your organisation only as well as it has been written down.
Layer three is a folder, not a technology project
Eight plain documents and about an hour: what a starter set looks like, so you can recognise one or notice it is missing.
What would this need to know?
The question to ask of any task before anyone aims a tool at it. When nobody can find the information, it is not an AI problem.
1 quiz
Section 8
Context is the job
What you will be able to do: Give a tool the context only you have, argue with its first draft, and make sure what you work out does not stay with you.
The stranger and the friend
Why a tool gives you a generic answer, and what changes when it has the context only you have.
GCSE: Goal, Context, Source, Expectations
One plain framework for briefing a tool, used everywhere, with a bad, an acceptable and a good example of the same task.
Sense, think, decide, act
A plain way to map your own work, so you can point at the step that is genuinely slow rather than the one that is annoying.
Argue with it
Why the first draft is where the work starts. The questions that test what came back, and why anything you type twice belongs in instructions or a file.
The thing you worked out should not stay with you
What you work out alone helps one person. Where organisations keep shared instructions and skills, who may add to them, and what it tells you when there is nowhere.
1 quiz
Section 9
Cultural sensitivity: whose knowledge is this?
What you will be able to do: Recognise material you hold but have no right to use, including Māori data and knowledge, and know who should ask and who decides.
Where ownership is not the test
Some material carries obligations that have nothing to do with who owns the file or whose system it sits in. Holding it is not the same as having the right to use it.
Whose data is it? Māori data and taonga
Why Māori data is taonga, why data sovereignty is about authority rather than where data is stored, and why an AI tool is not an authority on tikanga, mātauranga or te reo.
The relationship, not the compliance box
Asking the client or partner directly and early, rather than routing the question to the most senior Māori person in your building. The organisation owns the decision to ask.
What te Tiriti asks
Te Tiriti o Waitangi from its own text: what each article provides, the principles drawn from it, and what it asks of an organisation using these tools. Honest about where the guidance is still forming.
1 quiz · 1 practical exercise
Section 10
The value you bring
What you will be able to do: Separate the work you can hand over from the judgement that is your job, see what these tools can give back to people working around a barrier, and face honestly what happens to work that goes.
Delegate the repetition, keep the judgement
Every job is wrapped in paperwork nobody trained for. Handing that over is not about speed but about where your attention goes.
Critical thinking is the moat
When anyone can produce a fluent answer in seconds, judgement is the advantage. How to recognise the comfortable, well structured, generic first draft.
The tool that removes a barrier
For people working with dyslexia, ADHD, low vision, RSI, hearing loss or a second language, a tool can be the difference between doing something alone and having to ask. Nobody should have to disclose anything to benefit.
Roles deepen. Some tasks do not survive.
Tasks change, not whole jobs, but a job made mostly of those tasks is exposed. The honest version, and what actually helps.
1 quiz
Section 11
What to ask before you start
What you will be able to do: Ask the three questions that show how ready your organisation is, ask for a tool in a way an approver can act on, and know what to do and say when something looks wrong.
Your first three questions
Three questions that tell you almost everything about how ready your organisation is: where the source of truth is, where use is registered, and where problems get reported.
The Tool Case: five questions to answer before you ask
How to ask for an AI tool in a way an approver can act on: describe the job, not the product. And why assessing the vendor is not your job.
When something looks wrong, and what you tell people
Reporting what a tool gets wrong rather than working around it, owning mistakes early, and what to say when someone asks how the work was done.
1 quiz · 1 practical exercise