How to Use AI for Business Decisions Without Getting Generic Answers
Type "give me a SWOT analysis for my business" into an AI assistant and you'll get four boxes filled with things that would apply to almost any company: "strong brand," "limited marketing budget," "growing market," "economic uncertainty." None of it is false. It's also not useful, because it isn't actually about your business — it's about the average business the model imagines when you give it no real information.
Business strategy prompts fail for a specific, fixable reason: they ask for judgment without providing the facts judgment requires.
The core problem: strategy without specifics
A SWOT analysis, a pricing decision, a hiring call — every one of these requires context to answer well. A human consultant would ask a dozen follow-up questions before venturing an opinion. Most people skip that step with AI and expect a sharp answer anyway, so the model quietly fills in the blanks with generic business-school language instead of asking. The output isn't wrong so much as untethered — it could describe a hundred different companies equally well, which means it's not really describing yours.
Fix 1: Front-load the context a consultant would ask for
Before asking for the analysis, give the model what a real advisor would need: company size, current revenue stage, the specific market, what's already been tried, and the actual decision on the table. "I run a 6-person e-commerce brand doing $40k/month in candles, considering whether to launch a wholesale channel with 3 interested retailers already" gives the model something to actually reason about. "Should I expand my product line" gives it nothing to anchor to.
Fix 2: Ask for judgment, not just a categorized list
A SWOT analysis without a recommendation is homework, not advice. Explicitly ask the model to end with a stance: which single factor should actually decide this, and why. Forcing a model to commit to a specific recommendation — rather than neutrally listing pros and cons — pushes it to weigh the inputs you gave it instead of listing generic categories side by side with no synthesis.
Fix 3: Ban business-school stock phrases explicitly
Certain phrases are so overused in strategy content that they signal the model is coasting on training-data averages rather than reasoning from your specifics:
- "leverage synergies"
- "first-mover advantage" (unless genuinely true)
- "low-hanging fruit"
- "in today's competitive landscape"
- "think outside the box"
- "move the needle"
A single line banning these forces more concrete, situation-specific language in their place — the same trick that works for copywriting works here too.
Fix 4: Ask it to flag what it doesn't know
The most useful instruction you can add to any strategy prompt is permission to say "I don't have enough information to answer this part confidently." Left unprompted, a model will often produce a confident-sounding answer even when it's missing a critical fact, because confident answers are what most training data rewards. Explicitly telling it to flag gaps or state assumptions turns a plausible-sounding guess into something you can actually trust or correct.
Before vs. after
Weak prompt: "Give me a SWOT analysis for my consulting business."
Strengthened prompt: "I run a 3-person consulting firm specializing in HR compliance for companies under 50 employees, currently doing $15k/month, mostly from referrals with no outbound marketing. I'm considering hiring a fourth consultant to take on more clients. Give me a SWOT analysis with 3 specific points per quadrant grounded in this context, not generic business truisms. End with a one-paragraph recommendation on the single biggest factor that should decide whether to hire now, and flag any assumption you're making due to missing information."
The second prompt can't produce a generic answer even if it tried — every quadrant has to reference a 3-person firm, referral-based growth, and a specific hiring decision, which is what makes the output something you could actually act on.
If you'd rather not retype this structure every time, the Prompt Library has a ready-made SWOT analysis prompt and an OKR-planning prompt in the Business & Productivity category built around this same context-first approach.