Level 1 · Ask — Chat
Ask one expert
One expert who’s read almost everything. It reasons, you verify, you act.
- meeting-notes.docxWord · 3 pagesYou: Turn these notes into a one-page brief for Monday’s meeting.
- Claude:Read meeting-notes.docx
Done — the brief is on the right. Two decisions still need an owner.
Reply to Claude…
01 Equipment
Know what’s in the window
Chat has no hands on your accounts. Everything it uses, you load into this one window.
- 1
Sidebar & Projects
A Project gives every new chat the same files and instructions.
- 2
Upload a file
Grounds the answer in your document, not general knowledge.
- 3
Web search
Pulls current public facts with links — open them to check.
- 4
The artifact panel
The deliverable opens beside the chat. Copying it out is on you.
- 5
Composer & model
Pick a stronger model for harder thinking.
- 2q1-actuals.xlsxSheet · 4 tabsYou: Where did we overspend in Q1? Check current energy prices too.
- Claude:Read q1-actuals.xlsxSearched the web
Two lines stand out — a summary is on the right:
- Energy +18% vs budget
- Contractors +9%
Reply to Claude…
Simplified mock-up · example conversation
02 Try it
Fix a lazy prompt
Half of “bad answers” are really a bad ask. Tap a card and watch the same request sharpen.
Write something about our project update.
Write a 3-sentence update for our client (non-technical): the data migration is 2 weeks behind on vendor sign-off; training is on track. Keep it reassuring.
✗ No reader, tone, length or facts.
✓ Audience, length, facts and tone — all set.
What do you think about these numbers?
Here are Q1 energy costs vs budget [table]. Name the top 2 variances, a likely cause and one action each. Under 150 words, for a board pack.
✗ Which numbers? What decision?
✓ Data, task, format, limit, audience — locked in.
Summarize this.
Summarize this transcript in 5 bullets — decisions made and who owns each action. Ignore the chit-chat.
✗ How long? What angle? For whom?
✓ Format, length, focus and what to drop.
03 The shape of a good ask
Anatomy of a good ask
One real prompt, five parts. Leave one out and Claude fills the gap with a guess — usually a generic one.
- You’re a commercial analyst who lives in spreadsheets.Rolewho Claude is
- Tell me where we overspent in Q1 and what to do about it.Objectivewhat done is
- Actuals vs budget are attached; energy and contractors look high.Contextwhat only you know
- Under 150 words — it’s going in a board pack.Constraintsthe limits
- Give the top 2 variances, a likely cause and one action each.Outputthe shape
+ add files · / commandsEvery part removes a guess. Next: build your own, field by field.
04 Try it
Build it with ROCCO
Five fields: Role, Objective, Context, Constraints, Output. Fill them and the prompt writes itself.
Start from a preset
Load one, edit a field — the prompt rebuilds live.
Copy it into Claude
Run it and compare with your lazy version.
05 The verify reflex
Confident isn’t correct
Claude is always fluent. Accuracy is your job — and facts and judgments are checked opposite ways.
Facts → check outside the chat
Numbers, dates, citations. Look them up yourself.
Judgments → stress-test
Ask what has to be true, and when it wouldn’t hold.
Never let it cite itself
A made-up stat just gets a made-up source.
You should reforecast every 2 weeks. SaaS firms reforecast every 14 days on average, and a rolling forecast beats an annual budget when revenue is volatile.
Each claim below — how would you check it?
SaaS firms reforecast every 14 days on average⚠ hallucination
A precise, sourceless number — the classic tell. Search it and you’ll find nothing. Check facts outside the chat; never let the model cite itself.
A rolling forecast beats an annual budget when revenue is volatile
A recommendation, not a lookup. Stress-test it: what has to be true for this to hold — and when would it not?
You should reforecast every 2 weeks
Advice tuned to nothing about you. Pressure-test it against your actual situation before you act.
06 Your job at every level
Wear three hats
You carry these up the whole tower. At Level 1 the Verifier leads — mistakes are cheap here, so build the reflex now.
Navigator
You own the goal. One clause of context changes the whole answer.
Frame before you fire.
Verifier
You demand evidence. Fluency is constant; accuracy isn’t.
Test the reply as a claim.
Conductor
You steer across turns instead of taking the first reply.
Refine on purpose.
07 How careful to be
Match caution to what a wrong answer costs
Ask: if this is confidently wrong and I act on it, what does it cost?
Low — reversible
Move fast, light context, take the first good answer.
Medium — feeds a decision
Invest context, ask for the reasoning, spot-check the facts.
High — money, legal, published
Demand sources and verify everything. It’s a draft, never the authority.
08 Do it now
Four small steps, in real Claude
-
Step 1 of 4
Fix a vague prompt
Half of “bad answers” are really a bad ask.
You should see: The same request, reshaped so the answer can’t miss.
Use the Fix a prompt cards above — feel a lazy prompt become a sharp one. -
Step 2 of 4
Build with ROCCO
Role · Objective · Context · Constraints · Output.
You should see: A fully dressed prompt — and a much sharper reply.
In the ROCCO builder: load a preset, edit a field, copy the prompt and run it in Claude. -
Step 3 of 4
Make it show its work
Confident isn’t correct.
You should see: A claim you can test, not just trust.
paste into ClaudeAnswer, then name the one assumption that, if wrong, most changes your answer — and how I'd check it.
-
Step 4 of 4
Ship an artifact
If nothing left the chat, you only talked.
You should see: Something you can paste anywhere.
paste into ClaudeTurn this into a one-page brief: title, 3-bullet summary, one next step.
One step = one paste into Claude (the mobile app is perfect). Each step you mark done climbs a floor; progress saves on this device.
09 Decide
Is chat the right level for this?
You want Claude’s help right now. Which fits?
Pick the closest match — you’ll get a verdict.
10 Capstone
The whole level in one prompt
- Fill the brackets — they’re the five ROCCO fields.
- Paste it into Claude and read the reasoning.
- Push back on its weakest point, then check the shakiest fact outside the chat.
Role: [who Claude should be]. Objective: [the outcome, one line]. Context: [2–3 facts only you know]. Constraints: [limits — length, tone, what to avoid]. Output: [the shape you need]. Then: show your reasoning, flag general advice vs specific to me, and name the assumption most likely to be wrong. I'll push back on your weakest point.
Wrap-up · check yourself
You’ve got Ask
Remember
You’re ready when you frame before you fire, verify by type — and keep wishing Claude could just see and act on your real stuff. That wish is Level 2.