|
Hey there,
Week 3 of the AI, Eh? Summer Boot Camp. This week is AI for Work, without the jargon. Whether you work in an office, a shop, or run your own thing, one problem is universal: messy back and forth communication that buries the actual point.
Quick reminder of the shape, same every week. A detailed prompt you copy and paste. An iteration showing you how to fix it. Something personal you apply it to. And a Prompt Card you save to your toolkit.
Grab your toolkit doc. We are adding Card 3. Let's go.
Six things every good prompt needs
Same checklist as every week. Watch for all six in the prompt below.
|
1. Give it context. Who you are and what is actually going on, anything AI could not otherwise know. |
|
2. Be specific about the task. Not "help me with this email," exactly what you want it to produce. |
|
3. Set your real constraints. Tone, length, whatever actually limits your options. |
|
4. Tell it the format you want back. A list, a table, short sections, a word limit. Do not let it guess. |
|
5. Ask it to flag uncertainty. Anything it is not fully sure about, so you know what to double check yourself. |
|
6. Protect your privacy. Skip anything that could identify you personally, like your full name or address. |
Step 1: Run this prompt exactly as written
This one includes a sample email thread pasted right into it, so the whole thing is a little longer than usual. Open ChatGPT, Claude, or whatever AI tool you already have. Paste the entire block below in exactly as written, and hit enter.
"Here's a work email thread. Summarize it into a clear list of what needs to happen, who's responsible for each thing, and anything still unresolved. Keep it to short bullet points, plain language, no corporate phrasing. Flag anything unclear, contradictory, or left hanging, so I know what to follow up on before the day arrives.
From: Dave
To: Priya, Mark
Subject: Client visit Thursday
Just confirming Thursday's visit is still on for 10am. Priya, can you book the meeting room? Also, should we order lunch or just do coffee? And does anyone know if the client needs parking passes, I don't want to forget that again.
From: Priya
To: Dave, Mark
Subject: Re: Client visit Thursday
Room's booked. Heads up though, I've got a dentist appointment Thursday morning and might not be back until 11. Mark, could you cover the intro if I'm not there yet?
On lunch, might be worth checking budget first before we commit. I'll ask about parking passes when I confirm the room.
From: Mark
To: Dave, Priya
Subject: Re: Client visit Thursday
Sure, I can cover if needed. What time is the client actually landing though? I thought it was 11, not 10.
Also I don't have the budget numbers, Dave do you have those? If not I can dig them up tomorrow before this goes further.
From: Dave
To: Priya, Mark
Subject: Re: Client visit Thursday
Good catch, let me double check the actual arrival time and get back to you both. I'll also track down last quarter's event budget tonight so Mark doesn't have to." |
Step 2: Here is how you fix it when it is not quite right
You will probably get a clean, confident looking summary back. Look closely at how it handles the meeting time though. The thread never actually settles whether it is 10am or 11am, Dave said one thing, Mark thought another, and it was never resolved by the end. AI tends to smooth things like this over rather than flag them, quietly picking one time and presenting it as settled when it was not.
This is exactly the moment to speak up, in one line, in the same conversation.
| "The summary picked 10am as the meeting time, but the thread never actually settles that, Mark thought it was 11. Can you flag that as unresolved instead of choosing one." |
One sentence, pointed at the exact gap. This is a genuinely important habit at work, always compare a summary back against what was actually said, especially anything with a time, a number, or a commitment attached to it. AI is not being careless, it is just built to sound complete, even when the source material is not.
Step 3: Build your own, then save it as a Prompt Card
Now think of something real from your own job this week, a messy thread, a jumble of notes, a string of texts about a shift, whatever applies. In your head, walk through it. What is actually going on. What you actually want back. Your real limits, tone, length, whatever is boxing you in. How you want it formatted. What you want it to double check before you trust it. And keep it anonymous, real details, no names, companies, or anything identifying.
Now say all of that out loud as one real paragraph, paste your actual messy content underneath it, and run it.
Once you have a result, ask yourself two quick things. Did anything come back not quite right? If so, what one line did you send to fix it?
Now open your toolkit doc, add a heading called Prompt Card 3, and write out these four things yourself, based on what you just did.
|
What this is for. One line, plain language, describing your real situation. |
|
The prompt. The real paragraph you wrote and ran. |
|
The fix that got it there. The one line you sent to correct it, if you needed one. |
|
When to use it again. One line reminding future you when this applies. |
That is Prompt Card 3, done. Three real, working prompts in your toolkit now, each one proven on your own life.
Coming up next week
Week 4 is Learn Something New. Same four steps, pointed at anything you have been meaning to figure out. Bring your toolkit doc, Prompt Card 4 is next.
|
AI assisted, Human led.
Whatever AI produces for you, always read it, verify it, and make it sound like you. AI is an incredible first draft machine. Your judgment, your voice, and your values are always the final layer.
|
Hit reply and tell me what your Prompt Card 3 turned out to be. I read every single one.
Know someone who keeps saying they want to "get into AI" but does not know where to start? Forward this to them. Boot Camp starts wherever they join in.
Talk soon,
Chris
Founder, AI, Eh?
theaieh.ca π
|
