Prompts, Chatbots & Copilots
How employees actually use AI tools at work, and how to write prompts that get useful results safely.
What you'll learn
- Tell a chatbot apart from a copilot
- Write a clear prompt that gets a useful result
- Avoid pasting confidential data into the wrong tool
Most people meet AI at work through a chat box. You type a request in plain language, and a moment later you get back a draft email, a summary, a rewrite, or an explanation. It feels like magic, but the quality of what you get back depends heavily on what you put in and which tool you’re using. A little fluency here pays off every single day, because these tools touch the small writing-and-thinking tasks that fill an ordinary working week.
Chatbot, copilot, prompt
Three words come up constantly. A prompt is simply the instruction you type — your request in everyday language. A chatbot is a standalone tool you open in a browser or app and talk to directly; it knows general things but nothing about your specific work unless you tell it. A copilot is AI built into the software you already use — your email, your documents, your spreadsheets — so it can act on what’s in front of you. Ask a copilot to “summarise this thread” and it can see the thread; a generic chatbot cannot unless you paste it in.
A good prompt in, a draft out — and a human review before it counts.
Writing a prompt that works
The biggest improvement most people can make is to stop typing tiny requests and start giving context. A vague prompt like “write an email about the delay” gets a vague email. A strong prompt sets the role, the goal, the audience, and the format: “You’re helping me write a polite, brief email to a client explaining that delivery is delayed by three days, apologising, and offering a new date. Keep it under 120 words.” The model isn’t reading your mind; it’s responding to what you said. The more you tell it about tone, length, and who’s reading, the closer the first draft lands. A useful trick is to give an example of the style you want, or to paste in a previous message you were happy with and say “match this tone” — showing is often quicker than describing.
Treat the AI like a fast, eager junior colleague: brilliant at first drafts, but it needs clear instructions and always needs checking before anything goes out with your name on it.
Iterate, don’t expect perfection
The first answer is a starting point, not a finished product. Reply with adjustments — “make it warmer,” “cut the second paragraph,” “add a line about the refund policy.” This back-and-forth is normal and is where the real value sits. You bring the judgement about what’s right for your situation; the tool brings speed. Don’t be shy about throwing away an answer and starting over with a clearer prompt, either — it costs you nothing and often beats wrestling a weak draft into shape.
The one habit that keeps you safe
Be deliberate about what you paste in and which tool you use. Public consumer chatbots are not the place for confidential customer data, unreleased financials, personal information, or anything covered by a contract or privacy rule. Many companies provide an approved, enterprise version where your data isn’t used to train the model and stays within agreed boundaries — use that for work content. If you’re unsure whether a tool is approved, ask before you paste. A leaked customer list is a far bigger problem than a slow email.
Spot it: chatbot, copilot, or prompt issue?
Read each situation and decide for yourself, then tap a card to flip it and check your answer.
Sort the prompts
Drag each prompt into the bucket that best describes it — or tap an item, then tap a bucket. Hit Check placement when you’re done.
Here's where each one goes:
- Polite 100-word apology, warm but professional → Strong prompt — it sets tone, length, purpose, and audience clearly.
- "Write something about the project" → Weak prompt — gives the model almost no direction; the output will be equally vague.
- Rewrite to friendlier, under 60 words, with a tone example → Strong prompt — showing an example and constraining length are exactly the right moves.
- "Make it better" → Weak prompt — "better" means nothing without criteria; specify what needs to change and why.
- Three risks for a non-technical manager, one sentence each → Strong prompt — role, audience, and format are all specified.
- "Summarise everything" → Weak prompt — no length target, no audience, no focus — the summary will likely miss what matters.
Tip: drag with a mouse, or tap an item then tap a bucket on touch screens. Get one wrong and the answer key appears.
How to use it
Pick one routine writing task this week and hand the first draft to AI. Use a copilot when the content already lives in your tools, and a chatbot for general help. Try prompts like: “Summarise this report in five bullet points for a busy manager,” or “Rewrite this message to sound friendlier but still professional,” or “List three risks I might have missed in this plan.” Then read carefully, edit to fit, and confirm any facts. And before you paste anything sensitive, check that the tool is one your company has approved.
Quick check
1. The difference between a copilot and a generic chatbot is that a copilot…
2. A strong prompt usually includes…
3. Before pasting confidential customer data into a chatbot, you should…