Skip to main content
    Resources

    Guides and prompts, freely available

    A concrete glimpse of what we work on in training — usable right now

    Writing a good prompt for visual creation

    Whether for Midjourney, DALL·E or Sora, a good prompt follows the same logic: subject, style, composition, light, technical details. The more precise the description, the less the tool has to guess.

    Weak prompt

    “A photo of a modern office”

    Worked prompt

    “Wide-angle photo of a minimalist modern office, soft natural light from a bay window, beige and light-wood palette, a green plant in the foreground, architectural-photography style, 16:9 format”

    Three habits to build: iterate rather than cramming everything into one prompt, use precise style vocabulary (photorealistic, illustration, watercolour, viewing angle, lighting), and always state the expected output format (ratio, orientation, intended use).

    How to spot a task to automate

    Not all tasks are equal for automation. A good candidate usually ticks three boxes: it is repetitive, it follows clear rules (few exceptions), and it has an identifiable trigger (an email received, a new row in a spreadsheet, a recurring deadline).

    1. List the recurring tasks of the week or month.
    2. Estimate the time spent on each.
    3. Spot the ones with a clear trigger and little human judgement needed.
    4. Prototype with a simple tool before scaling up.

    Common examples: sorting and categorising emails, summarising meeting notes, extracting data from documents, answering FAQs, drafting recurring content. Conversely, keep a human in control of any sensitive or high-stakes decision — that's the very principle of the responsible framework covered in training.

    Office prompt library

    For Copilot (Microsoft 365) or Gemini (Google Workspace) — click to copy

    Word / Google Docs

    Excel / Google Sheets

    Outlook / Gmail

    PowerPoint / Slides

    Want to go further?

    These resources are a glimpse — training goes further, on your own use cases