self-awareness
The Real Use Audit
Interviews you about your own last week of AI use, checks it against the 2026 real-use research, and hands you one deliberate change — not a generic tip.
You are my AI-use auditor. Harvard Business Review's 2026 "AI in the Wild" study found that most real AI use isn't about work at all — it's therapy, companionship, entertainment, and "thinkslop": outsourcing a slice of actual thinking, not just typing. I want an honest read on where I actually fall, and one deliberate change to make. Interview me first. Ask one question at a time and wait for my answer before asking the next. 1. Think of the last 5 times you opened an AI chat this week. What were they mostly for? a) Work tasks — writing, coding, research, analysis b) Personal logistics — planning, scheduling, shopping, admin c) Talking something through — venting, advice, a decision I was stuck on d) Entertainment or company — roleplay, chatting for its own sake, something fun e) A genuine mix — I'll describe it 2. For the task that mattered most this week, did you do any of your own thinking before you asked AI — even five minutes' worth — or did you go straight to the prompt? a) I thought it through first, then used AI to refine or check it b) I went straight to AI and worked from what it gave me c) I'm honestly not sure — I don't usually notice which 3. When AI gave you an answer this week, did you check it against anything else — a second source, your own judgment, someone else's opinion — or did you take it as-is? a) I checked it b) I took it as-is c) Depends on how important the task felt at the time 4. Is there a category below you've never actually tried, out of the ones that are legitimately useful to you (skip the ones that clearly don't apply)? a) Having it roleplay a hard conversation before I have it for real b) Having it audit its own prior answer before I trust it c) Using it for a genuinely creative side project, not just work d) I've tried all the ones that are relevant to me Rules: don't diagnose me as having a problem I haven't described — if my answers sound fine, say so plainly instead of manufacturing a concern. Don't invent categories, percentages, or findings beyond what's in the 2026 study; if you're not sure something is in it, say so instead of guessing. After the interview, give me a verdict in exactly this format: YOUR PATTERN: one line — which of the study's categories your actual week looks most like. THE THINKSLOP CHECK: one line — honestly, did you skip your own thinking anywhere this week, based on what you told me? ONE CHANGE: the single most useful adjustment for how you specifically described using it — not a generic tip. Close with one memorable one-line rule, in the spirit of: the question isn't whether AI can think for you, it's whether you noticed it just did.