Business Decision Framework

This is the prompt template Daniel used in the StackSensible side-by-side test of Claude vs ChatGPT on a real $51,200 vendor consolidation decision. Paste it as-is into either tool, fill in your own scenario, and run the steelman follow-up at the end.

The prompt

You are a senior operations advisor helping a small business owner make a vendor consolidation decision.

SCENARIO
Total annual spend across category: $[TOTAL ANNUAL SPEND]
Number of current vendors: [N]
Category: [WHAT YOU BUY — e.g. SaaS, manufacturing inputs, professional services]
Business size: [HEADCOUNT + ANNUAL REVENUE]
Decision deadline: [DATE OR "FLEXIBLE"]

SUPPLIER DATA

Supplier A:
  - Annual cost at current volume: $[X]
  - Payment terms: [e.g. 30 days upfront / net-30 / net-60]
  - Contract length: [e.g. 12-month / month-to-month]
  - Exit clause: [e.g. 30-day notice / 90-day notice + $X penalty]
  - Notable terms: [SLAs, volume commitments, exclusivity, etc]

Supplier B:
  - Annual cost at current volume: $[Y]
  - Payment terms: [...]
  - Contract length: [...]
  - Exit clause: [...]
  - Notable terms: [...]

Supplier C:
  - Annual cost at current volume: $[Z]
  - Payment terms: [...]
  - Contract length: [...]
  - Exit clause: [...]
  - Notable terms: [...]

YOUR TASK
1. Identify the lowest unit-cost supplier on face value.
2. Identify all material constraints I should consider beyond unit cost (contract risk, cash flow risk, operational risk, concentration risk).
3. Recommend a consolidation strategy with a clear primary recommendation.
4. Label every assumption you're making about my business that could change your recommendation if wrong.
5. State the conditions under which you would revisit the recommendation.

Be specific. Show your reasoning. If you don't have enough information for a step, say what you'd need.

The steelman follow-up

After Claude / ChatGPT has given you their recommendation, send this as your next message in the same conversation:

What is the strongest argument AGAINST your recommendation? Steelman the opposing view. What would have to be true for that opposing view to win?

This is the test. Claude (in our experience) responds with a substantive counter-argument naming specific risks. ChatGPT typically restates the original recommendation with softer language. The difference is the whole point of the comparison.

What to look for in the output

Strong signals

• The tool surfaces a risk you hadn't explicitly named (e.g. exit clause, concentration risk, payment-term cash flow impact)
• It labels its assumptions and tells you what would change its mind
• The steelman is substantive — names specific risks, not just "some stakeholders may prefer"
• It cites the specific clauses or data you provided, not generic industry benchmarks

Weak signals (means: don't trust this output for high-stakes decisions)

• Confident percentage savings figure with no source ("typical industry savings of 18%" without you supplying that number)
• The recommendation ignores a contract clause you provided in the input
• The steelman is a disclaimer ("while this has merit, some may prefer") rather than a counter-argument
• Output reads as polished marketing rather than analytical thinking

Methodology context

This prompt was used in the StackSensible video "Claude vs ChatGPT for Business — I gave both the same $51K decision". The video walks through the actual outputs both AIs gave to a real Australian SMB vendor consolidation scenario. The full methodology — including the vendor scenario, the specific risk Claude caught vs ChatGPT missed, and the cost-of-acting-on-wrong-output analysis — is at the methodology page.

For developers / power users

If you're building this into an agent workflow, this prompt is also available as a structured template at:

• Anthropic Claude → Prompt engineering best practices
• OpenAI → Prompt engineering guide