AI Guides › Claude Mastery
Claude For Planning A Project
By Nigel Guy · 2 min read
Ask Claude to "help me plan this project" and you'll get a plan — cheerful,
plausible, and often useless, because a goal isn't a plan and most requests
never get more specific than the goal. The plan reads well and doesn't
survive contact with the actual first Tuesday.
The rule: a real plan needs a sequence, an owner, and a definition of done
for each step — a bullet list of good intentions isn't one of those things.
The mechanism
- State the actual constraint, not just the goal. "Launch this by March
3rd with one other person helping" produces a different, more honest plan
than "help me launch this."
- Ask for the plan in dependency order, not priority order. What has to
happen before what — a task list ranked by importance still leaves you
guessing what to actually do Monday morning.
- Force a definition of "done" onto every step. "Draft the outreach
list" is vague; "outreach list of 30 named contacts with an email for
each" is checkable. If a step can't be checked off unambiguously, it's not
ready to go on the plan yet.
- Ask it to name the step most likely to slip, and why. This single
question surfaces the real risk in the plan far faster than reviewing the
whole thing line by line.
- Revisit the plan weekly, not once. A plan made once and never revised
is a wish list with dates on it.
What to skip
Skip asking for a plan before you've stated the constraint — you'll get
something generic and have to redo it anyway. And skip demanding a fully
detailed 90-day plan on day one; a solid next-two-weeks plan you'll actually
follow beats an impressive quarter-long one you'll abandon by week three.
Guardrails
- A plan is a starting hypothesis, not a commitment device. Treat the first
version as something to pressure-test, not something to defend.
- Anything with a hard external dependency (someone else's approval, a
vendor's timeline) needs that dependency named explicitly — a plan that
assumes everyone else moves on your schedule isn't a plan, it's a hope.
- The step most likely to slip is usually the one nobody wanted to name out
loud. Ask for it directly rather than waiting to discover it in week three.
All 751 AI guides · JulieMango plans from £17/mo