Phase 1 · ChatGPT · Level 1 · Foundations
ChatGPT: Foundations Quiz · 10 min
Why it matters A quick check of your judgement before you move on. There's no penalty for a low score, and you can retake it, or take it first to test out of a level you already know.
Nicely done getting through Level 1. These aren't trivia; they're the everyday judgement calls you'll face using ChatGPT well and safely at work. Pick the best option for each scenario, and you'll see an explanation after each one.
Passing this checkpoint: work through the level first, then score 70% or more here.
Skipping this level: already confident? Take this cold and score 80% or more to test out and jump ahead. Below that, nothing is lost, you just study the level as normal.
1. You ask ChatGPT to help write a work report and it gives you a confident paragraph with a specific statistic and a named study as the source. What's the wisest next step? Paste it straight into the report; it named a study, so it must be real. Find the named study yourself and confirm both it and the statistic are genuine before using them. Ask ChatGPT 'are you sure that study is real?' and trust it if it says yes. Remove the statistic but keep the study name, since names are usually accurate. 2. You've spent a long chat getting a client proposal just right. Now you need help with a completely separate task, a quick internal rota. What's the tidiest approach? Carry on in the same chat; ChatGPT keeps unrelated topics neatly apart. Click 'New chat' and start the rota in a fresh conversation. Delete the proposal chat first, then reuse it for the rota. Open a second ChatGPT account so the two never mix. 3. Your first draft of an email from ChatGPT is fine but a bit long and formal for a colleague you know well. What gets the best result? Accept it as-is; the first answer is usually the intended one. Start a brand-new chat and describe the whole email again. Stay in the same chat and nudge it: 'shorter, please,' then 'a bit warmer, we're on friendly terms.' Give up on ChatGPT for emails and write it from scratch yourself. 4. You need to know who currently holds a particular public role, and when they were appointed, for something you're publishing today. What's the safest approach? Trust ChatGPT's built-in answer; it's good with facts like this. Use its web search or check an official website, because its trained-in knowledge may be out of date. Ask ChatGPT and, if it sounds confident, assume it's current. Assume ChatGPT can never help with anything factual and don't bother. 5. You're about to paste a customer's complaint email into your personal ChatGPT account to draft a reply. It contains their full name, address and account number. What should you do first? Paste it all in; you need the reply quickly. Remove or replace the personal details and give ChatGPT just the situation. Paste it in but ask ChatGPT to keep it private and forget it afterwards. Nothing; customer details are fine to share with a personal AI account. 6. You're new to ChatGPT and haven't hit any usage limits. A pop-up invites you to upgrade to a paid plan. What's the sensible move? Upgrade straight away so you get the best possible answers. Stay on the free version until you hit a specific limit that's actually getting in your way. Upgrade so it's then safe to paste in confidential work documents. Upgrade because paid answers are always more accurate. 7. You have two jobs for ChatGPT: quickly rewrite a one-line email to sound friendlier, and work out the best meeting slot for four people with clashing constraints. How should you use the model picker? Use the slower 'thinking' model for both, to be safe. Use the fast everyday model for the email rewrite, and the slower thinking model for the scheduling puzzle. Use the fast everyday model for both; the picker makes no real difference. Avoid the picker entirely and never change it from whatever's selected. 8. After a workshop, Maya photographs the whiteboard (including some cost figures) and asks ChatGPT to type it up as notes. The result looks clean and complete. What should she do before sending it round? Send it as-is; image understanding reads photos accurately. Check the typed figures against the actual whiteboard, especially the numbers, before trusting them. Assume any hard-to-read parts were left out, since ChatGPT never guesses. Retake the photo and re-run it until the answer changes, then use that version. 9. You ask ChatGPT to turn the weekly meeting notes into an action table. Two jobs in the notes clearly have no owner, but the table confidently assigns them to named colleagues. What's the right response? Accept it; ChatGPT must have inferred the right owners from context. Tell it those two have no owner in the notes, have it mark them UNASSIGNED, and flag them as needing an owner. Delete those two rows so the table looks complete. Send the table round and let the named people object if they disagree. Answered 0 of 9.
Answer every question to see your score