ChatGPT for Business Analysts: 30 Prompts You'll Use Daily

In my BA practice, I see analysts who treat ChatGPT as a search engine missing most of its value. Treated as a structured collaborator with the right prompts, ChatGPT (or Claude, or Gemini) has become the most-used tool in my day. In this guide I share 30 prompts organised by BA workflow, each one I’ve tested against real engagements.

Strong BA prompts have four parts:

• Role: who the AI acts as.
• Goal: specific outcome.
• Context: what the AI needs to know.
• Output format: how the response should be structured.

Without all four, prompts return generic output. With all four, output is reliable.

P1: Interview guide > “I am interviewing the [role] about [topic]. Generate a 45-minute interview guide. Include rapport-building, current-state, future-state, constraints, success criteria. Open-ended, no leading questions.”

P2: Probing questions > “When the stakeholder says ‘[paste statement]’, generate 4 follow-up questions that probe specifics, impact, alternatives tried, and ideal solution.”

P3: Stakeholder-specific guide > “Take this base interview guide and rewrite for a CFO audience. Adjust language and examples for senior finance perspective.”

P4: Asynchronous elicitation > “Convert this interview guide to an async survey format. 12 questions max. Mix of open-ended and structured.”

P5: Workshop facilitation prep > “I am facilitating a 2-hour requirements workshop with 8 stakeholders. Generate an agenda with timings, suggested activities, expected outputs.”

P6: Theme clustering > “Below are 18 interview transcripts. Cluster requirements into 8-10 themes. For each: name, frequency, dominant viewpoint, contested points, supporting quotes.”

P7: Surprise finder > “Of these synthesis themes, what is most surprising or counter-intuitive? What are the implications? What further research would confirm?”

P8: Contradiction detector > “Read these stakeholder inputs. Identify any contradictions across stakeholders. For each: the contradiction, who said what, suggested resolution path.”

P9: Gap analyser > “Below are extracted requirements. Compare against standard categories: security, performance, scalability, accessibility, compliance. Flag categories that are unaddressed.”

P10: Quote pull > “From these transcripts, pull 5 powerful quotes that an executive could use to support investment in [area]. Bias toward emotional impact and specific business consequences.”

P11: BRD draft > “From these synthesised themes, generate a BRD with sections: business context, scope, functional requirements, NFRs, assumptions, constraints, dependencies, AC, glossary. Tone: precise. Length: 4,000 words.”

P12: Use case generator > “Take this requirement and generate a use case. Include: actor, preconditions, main flow, alternative flows, exception flows, postconditions.”

P13: Acceptance criteria > “Generate 8-12 AC for this requirement. Use Given/When/Then. Cover happy path, edge cases, error states, observability, DoD.”

P14: Audience tailoring > “Take this BRD and rewrite the executive summary for C-level audience. 1 page, business outcomes first, no jargon.”

P15: Glossary builder > “From this document, extract all jargon, acronyms, and technical terms. Build a glossary with plain-language definitions.”

P16: BPMN from prose > “Convert this process description into BPMN. List all activities, decisions, swimlanes. Output as structured text I can paste into Lucidchart.”

P17: As-is to to-be > “From this as-is process description and these improvement goals, generate a to-be process. Highlight the changes.”

P18: Process gap analysis > “Compare this as-is process to industry best practice. Identify gaps and suggest improvements.”

P19: Process metrics > “For this process, suggest 5-7 metrics to track effectiveness. Include leading and lagging indicators.”

P20: Stakeholder register > “From this org chart and project description, suggest a stakeholder register. For each: role, influence, interest, suggested engagement, key concerns.”

P21: Power-interest grid > “Place these stakeholders on a power-interest grid. Justify placement with evidence from the project description.”

P22: Engagement plan > “For each stakeholder in this register, suggest engagement frequency, channel, content type, and risk if not engaged.”

P23: RACI matrix > “Generate a RACI matrix for this project’s key activities. Use these stakeholders [list]. Justify R, A, C, I assignments.”

P24: Validation questions > “For each of these 30 requirements, generate one validation question that confirms the requirement matches stakeholder intent.”

P25: Ambiguity detector > “Below are 50 requirements. Identify any that are ambiguous, untestable, or open to multiple interpretations. Suggest specific clarifications.”

P26: Test scenario generator > “For this requirement, generate 5 test scenarios covering happy path, edge cases, error states.”

P27: Options analysis > “From this problem statement, generate 3-5 options. For each: description, cost (rough), benefits, risks, implementation complexity. Recommend with reasoning.”

P28: SWOT generator > “Generate a SWOT analysis for [organisation/project]. Cover strengths, weaknesses, opportunities, threats. 3-5 items per quadrant. Justify each with evidence.”

P29: Business case refresh > “Update this business case with these changes [paste]. Recompute NPV, payback, sensitivities. Update the recommendation if appropriate.”

P30: Value realisation tracker > “Define metrics that would prove this initiative delivered the intended value. For each: metric, target, measurement method, frequency, owner.”

Replace generic [paste] markers with your actual data.

• Add your team’s vocabulary in the role section.
• Specify output format clearly.
• Iterate the prompt 2-3 times to calibrate.

Once a prompt works, save it. Reuse it weekly.

• Vague role. “You are a BA” is weaker than “You are a senior BA at a B2B SaaS company specialising in financial services.”
• No output format. Without one, output is unstructured prose.
• Asking for opinions instead of analysis. AI gives generic opinions. Ask for analysis grounded in inputs.
• Skipping context. AI cannot read your situation. Provide enough context.
• Using AI for things requiring human judgement. AI assists thinking, not replaces it.

Use this six-step process for any new prompt:

Step 1: Identify the recurring task (more than 30 minutes weekly).

Step 2: Define inputs and desired output format.

Step 3: Draft using the four-part pattern (role, goal, context, format).

Step 4: Test with real data. Iterate 3-5 times until output is consistently usable.

Step 5: Save with clear title and use case.

Step 6: Share with team and refine collectively.

Most BAs develop 15-20 personal prompts within 6 months of regular AI use.

Treat prompts like code:

• Save in a versioned library.
• Note which model and version produced best results.
• Refresh quarterly when models update.
• Retire prompts that no longer produce value.
• Document edge cases that broke the prompt.

A maintained prompt library is a multi-year career asset.