Free workshop - May 26, 2026

3 concrete ways to use AI at work

Write faster, create quality visuals, and turn your information into clear actions. Keep this page open during the workshop to copy prompts and open useful links.

Three use cases you can apply directly at work.

The session stays practical: writing, creating visuals, and turning your own information into next actions.

01

Wispr Flow

Write, dictate, and respond up to 4x faster.

02

ChatGPT Image 2.0

Create visuals, ads, and concepts without a heavy creative cycle.

03

ChatGPT with your real information

Turn notes, emails, and documents into clear actions.

4 prompts worth keeping

The four workshop prompts that still earn their place every week, rewritten to guide the assistant better. De-identify your text before pasting it.

Prompt 1

Create my custom instructions

I'm setting up an AI assistant for my work and I want to write my custom instructions: the short text, saved in the settings, that the assistant reads at the start of every conversation. With good instructions, I no longer have to repeat who I am or how I like to work.

Run a short interview to gather what you need, then write them for me.

How to run the interview:
- Ask one question at a time and wait for my answer.
- Keep it to about eight questions. I'm new to this, so offer two or three example answers with each question so I can respond quickly.
- If my answer is vague, ask one follow-up question before moving on.
- Start by asking which assistant I use and what fields I see in its personalization settings. Each tool splits them differently.

What you need to learn:
- my role, my type of organization and the people I serve, in general terms;
- the three or four tasks I'll ask for most often;
- who usually reads what I write: clients, colleagues, the general public;
- the language and tone I want, including Canadian spelling and how formal I am with people;
- the shape of answer that helps me: length, structure, level of explanation;
- my level of experience in my field and with AI, so you know what to explain and what to skip;
- what annoys me in answers from AI assistants;
- what the assistant should do when my request is ambiguous or when it isn't sure of a fact.

To capture how I write, also ask me to paste one or two short texts I wrote myself, with names and confidential details removed. I can skip this step.

Don't ask me for personal information about other people, client names, passwords or confidential details. The instructions should describe my work in general terms only.

When you have enough, write the instructions:
- in the first person, in plain sentences, as if I were explaining how I work to a capable new colleague;
- with the reason behind a preference when it helps apply it well, for example: "I often read on my phone between meetings, so get to the point";
- with only what stays true in all my conversations. Details of a one-off task don't belong here;
- with these two working rules, unless I tell you otherwise: ask me a question when my request is ambiguous instead of guessing, and tell me plainly when a fact is uncertain or needs to be checked;
- in under 1,500 characters in total, so the text fits in any assistant. If my settings have several fields, split the text across them and name each one.

Then:
1. Show the instructions in a block that's ready to copy.
2. Give me a short request to try in a new conversation so I can see the difference.
3. Ask me what I would change, then revise until I'm satisfied.
Prompt 2

Find my first three uses

Help me find my first three uses for AI at work.

Ask one question at a time and wait for my answer. First have me describe my role, then a normal week: the tasks that keep coming back, the ones that take the most time and the ones I always put off. Five or six questions should be enough. Don't ask for client or employee names or confidential details.

Then sort my tasks into a table with three categories:
1. hand to AI, with my review;
2. do with AI, because my judgment matters at every step;
3. keep for myself, or put rules in place before touching: personal information, decisions that affect a person, commitments, legal or financial questions.

For each task, give the reason for the category in one sentence.

Then suggest the three tasks to start with. Choose ones that are frequent, easy to check and without serious consequences if the first result is imperfect. For each, give me:
- a first request that's ready to copy, where I only need to fill in the parts in brackets;
- what I need to give the assistant;
- what I need to check before using the result.

Stay concrete. If one of my tasks isn't a good fit for AI, just tell me.
Prompt 3

Meeting notes to action plan

Turn my meeting notes into an action plan. The notes are at the end of this message. They are de-identified: keep the roles and labels as they are, without trying to guess who is who.

Work only from what is written. Don't add any fact, date or owner that isn't in the notes. When information is missing, write "to confirm" instead of inferring it.

Give me, in this order:
1. The decisions made, one line each.
2. A table of actions with four columns: action, owner, deadline, concrete next step.
3. The questions left open and who could answer them.
4. The "to confirm" items, grouped, so I can settle them in a single message.
5. A short follow-up email to participants, clear and direct, that restates the decisions and actions. Leave in brackets anything I need to fill in.

If two parts of the notes contradict each other, point it out instead of choosing.

Notes:
[paste the de-identified notes]
Prompt 4

Compare quotes

Help me compare supplier quotes or offers before I make a decision. I'll attach or paste them after your first question. I've removed people's names and confidential details.

Before comparing, ask me what matters most for this decision: price, timeline, quality, service, risk or something else. One question only, then ask me for the documents.

Then build a comparison table: supplier, total price, what is included, what is excluded, timeline, payment terms, warranties and how long the offer is valid. For each amount and each condition, note where it appears in the document. If information is missing, write "not specified". Don't estimate or infer anything.

Then add:
1. The differences that prevent a direct comparison, for example a scope or quantities that aren't the same.
2. The unclear or unusual points in each offer.
3. The questions to ask each supplier before deciding.
4. What my priorities suggest, with the reasons, and what could change that reading.

The decision is mine. Present the analysis without choosing for me, and remind me what to check before signing.

Continue with AI 101: steps for ChatGPT, Claude and Copilot

AI in Quebec: 5 essential reflexes under Law 25

AI can save a lot of time, but a company remains responsible for the personal information it uses, shares, or entrusts to an external tool.

  1. 01

    Identify personal information

    Before using ChatGPT with a document, ask whether the document can identify a person directly or indirectly.

    Examples:
    • Name, email, phone number, or address.
    • Client number, financial data, HR file, or sensitive information.
  2. 02

    Share only the necessary information

    ChatGPT does not always need to know the client’s name, address, or full case details. The less data you send, the less risk you create.

  3. 03

    De-personalize before sending

    Before giving a document to AI, replace sensitive information with generic terms.

    Examples:
    • “Client A”, “manufacturing SMB”, “Quebec region”.
    • “invoice F-001” or “employee concerned”.
  4. 04

    Verify the tool being used

    Not all AI accounts and tools are equivalent. Before using real data, ask the right questions.

    Questions:
    • Is the tool approved by the company?
    • Where does the data go, who can access it, and is it used to train the model?
  5. 05

    Keep humans responsible

    AI can help draft, summarize, compare, and analyze. For important decisions, especially in HR, finance, legal, or client relationships: AI proposes, humans decide.

Golden rule

You can delegate work to AI, but you do not delegate responsibility.

Keep judgment, validation, and final responsibility inside the company, especially when people, clients, or sensitive decisions are involved.

Workshop resources

Open Wispr Flow using the link below.

After the session

Keep what works. Improve what gets stuck.

Capture the prompts that produce useful answers, the criteria that improve quality, and the questions that keep coming back. Those are the best candidates for team templates.