Reasoning prompts

The concept

For anything with moving parts — a pricing decision, a strategy, a diagnosis of why a campaign underperformed, a contract risk review — the biggest quality gain comes from making the model think before it answers. Left alone, a model tends to commit to a conclusion in its first sentence and then spend the rest of the response justifying it. If the first sentence was wrong, everything after is elaborate rationalisation. The fix is to force sequence: analyse first, conclude second. Modern reasoning models do a lot of this internally, but explicitly structuring the thinking still improves the answer measurably, and — just as importantly — it lets you see and audit the reasoning instead of taking a verdict on faith.

Three techniques, in increasing power:

1. Step-by-step. "Think through this step by step before answering." Cheap, one line, works on almost everything analytical. Useless on creative or purely stylistic tasks.

2. Decomposition. You define the steps rather than leaving them to the model. "First list the constraints. Then list the options. Then score each option against each constraint. Then recommend." This is dramatically better than generic step-by-step because you're injecting your professional judgement about what the right analysis is — which is the part you're actually good at.

3. Plan-before-answer. For long or complex deliverables: "Before writing, give me a one-paragraph plan: structure, what goes in each section, what you'll leave out. Wait for my approval." You correct a bad structure in 30 seconds instead of reading 1,500 wrong words. For any deliverable over about 800 words, this is the highest-return habit in this entire course.

A fourth, situational: force alternatives before recommending. "Give me three genuinely different approaches, including one you think is risky, then recommend one." Models are agreeable by default and will happily validate whatever direction you implied. Demanding a real spread of options is how you stop it from being a mirror.

Before → After

Should we take on the retainer at EGP 180K/month?

Help me decide on a retainer. Work in this exact order and do not
jump to the recommendation.

STEP 1 — Restate the deal in your own words, and list every assumption
you're making that I haven't stated. Flag the assumptions that would
change the answer if wrong.

STEP 2 — Build the cost side. Deliverables are: 12 social posts/month,
1 video/month (1 shoot day), monthly reporting, 1 strategy session.
Estimate the internal hours per role (creative, production, account,
post) and total monthly cost at EGP 400/hr blended. Show the table.

STEP 3 — List the non-financial factors both ways: what this client
does for our positioning, portfolio, and pipeline; and what it costs
in team capacity and opportunity.

STEP 4 — Give me three options: accept as-is, counter, decline. For
each: the case for it, and the single biggest risk.

STEP 5 — Now recommend one, in three sentences, and state what would
have to be true for you to change your mind.