brainbaby

AI Prompts for Business

Fifteen prompts for strategy, meetings, hiring, and analysis — built to produce arguments you can pressure-test, not consultant-speak.

Start with a clear task

Ask AI for business advice and you get the management-book average: know your customer, focus on value, iterate quickly. All true, all useless, because it applies to everyone.

The reason is that business questions are almost entirely context-dependent. The right answer for a 3-person bootstrapped company is the wrong answer for a 200-person funded one, and the model defaults to the middle unless you tell it where you actually are.

The prompts below force specificity in two ways: they supply real constraints, and several of them ask the model to argue against a position rather than for one — which is where the useful thinking lives.

The technique: make it argue the other side

The most valuable business prompts don't ask "what should I do?" — they ask the model to attack a decision you've already made. Models are agreeable by default; asking for validation reliably gets it, which tells you nothing.

Framing the request adversarially — pre-mortems, steelmanned counter-arguments, "what would have to be true" — converts a yes-machine into a stress test. This is the difference between AI as a mirror and AI as a sounding board.

  • Weak prompt

    Should I raise my prices?

  • Strong prompt

    I'm planning to raise prices from [X] to [Y] for [product], sold to [customer type]. Our context: [churn rate, competitive position, how long since last increase]. Run a pre-mortem: it's six months later and this went badly. Write the story of what happened. Then tell me which of those failure modes is most likely given our specifics, and what leading indicator would warn me early.
  • Why it works

    A pre-mortem produces concrete failure paths instead of generic caution. And asking for a leading indicator converts the analysis into something you can actually monitor.

The prompts

  • Pre-mortem

    We're about to [decision]. Context: [constraints, stage, resources]. It's [timeframe] later and this failed badly. Write the story of how. Then rank the failure modes by likelihood given our specifics, and give me an early warning indicator for the top one.
  • Steelman the opposite

    I believe [position] about [business decision]. Context: [details]. Make the strongest possible case AGAINST my position — not strawman objections, the argument a smart person who disagrees would actually make. Then tell me what evidence would settle it.
  • What would have to be true

    For [plan] to work, what would have to be true about our market, our customers, and our execution? List each assumption and mark which are (a) already validated, (b) testable cheaply, (c) expensive bets. We're at [stage] with [resources].
  • Meeting to decisions

    Here are my meeting notes: [paste]. Extract: decisions actually made (not topics discussed), owners, deadlines, and open questions that were raised but not resolved. Flag anything where the notes are ambiguous about who owns it.
  • Prioritization forcing function

    Here are the [N] things we could build next: [list]. Our constraints: [team size, timeline, strategic goal]. Force-rank them and tell me explicitly what we're giving up by not doing the ones at the bottom. No 'it depends'.
  • Pricing analysis

    We charge [price] for [product] to [segment]. Competitors charge [range]. Our differentiation is [X]. Analyze whether we're mispriced — argue both that we're too cheap and too expensive, then tell me which case is stronger and what data would resolve it.
  • Churn diagnosis

    Our churn is [rate] and here's what we know about who leaves: [paste]. Generate hypotheses for why, ranked by how much of the churn each would explain. For each, tell me the cheapest way to test it.
  • Hiring scorecard

    I'm hiring a [role] for a [company stage/size] company. The main problem they need to solve is [problem]. Write a scorecard: the 5 outcomes they must deliver in year one, and for each, an interview question that would actually predict it. No generic culture-fit questions.
  • Difficult conversation prep

    I need to tell [person/role] that [difficult message]. Context: [relationship, history, stakes]. Draft what I should say — direct, not cushioned into vagueness. Then list the three most likely responses and how I should handle each.
  • Competitive positioning

    We compete with [competitors]. Here's what each does better than us: [honest list]. Here's our real advantage: [X]. Given that, what market position can we credibly own — and which customers should we deliberately not pursue?
  • Process bottleneck

    Here's how [process] currently works: [describe each step and who's involved]. Find the actual bottleneck, not the most annoying step. Tell me what evidence supports your answer and what would falsify it.
  • Metric critique

    We track these metrics: [list]. For each, tell me: what behavior does this incentivize, and how could someone hit the number while making the business worse? Which metric is most gameable?
  • Investor question prep

    We're pitching [business] at [stage] with these numbers: [paste]. Play a skeptical investor. Ask me the 10 hardest questions — the ones about the weaknesses in these specific numbers, not generic due diligence.
  • Build vs buy

    We need [capability]. Building costs roughly [estimate] and [time]; buying costs [price]. Our context: [team, strategic importance, timeline]. Argue both sides, then tell me which way you'd go and what factor would flip your answer.
  • Post-mortem facilitation

    Here's what happened: [describe incident/failure]. Facilitate a blameless post-mortem: what was the sequence, where were the decision points, what made the wrong choice reasonable at the time, and what systemic change would prevent recurrence? No individual blame.

Common mistakes

  • Asking for validation

    "Is this a good idea?" gets you a yes. Ask it to argue against you instead — the disagreement is the value.

  • Omitting your constraints

    Without your stage, team size, and runway, you get advice calibrated to a company that doesn't exist. Specificity in equals specificity out.

  • Treating output as analysis

    The model has no data about your market. It structures thinking well; it doesn't know your numbers. Supply real data or treat conclusions as hypotheses.

  • Accepting 'it depends'

    If you get a hedged non-answer, add constraints and demand a ranked recommendation. Force a position you can then argue with.

Frequently asked questions

Can AI give good business advice?

It's useful for structuring decisions, running pre-mortems, and pressure-testing your reasoning. It's not useful as an oracle — it has no data about your market, customers, or numbers beyond what you give it.

Why does AI business advice sound so generic?

Because the question was generic. Business answers depend almost entirely on stage, resources, and market position — supply those and the output changes substantially.

What's the single most useful business prompt?

The pre-mortem: describe the decision, then ask it to write the story of how this failed six months from now. It surfaces concrete risks that a 'what are the risks?' question never produces.

Should I share confidential business data with AI?

Check your tool's data-handling terms first. Look specifically for whether inputs are used to train models and what the retention policy is — this varies significantly between providers and plans.

How do I stop it from just agreeing with me?

Explicitly ask it to argue the opposite position, or frame the prompt as a critique request. Models default to agreeable; the framing has to override that.