Copilot prompts for business analysts

Build a copy-paste prompt pack for me as a business analyst working on [PROJECT OR DOMAIN] for [STAKEHOLDERS]. My available source material is [SOURCE MATERIAL], and my current tools and limits are [TOOLS AND CONSTRAINTS].

Create 10 prompts I can use in ChatGPT, Claude, or Copilot. Each prompt must be written in first person as if I am sending it, contain its own bracketed inputs, and be 70-120 words. Cover these day-to-day tasks:
1. Turn stakeholder notes into testable business requirements and acceptance criteria.
2. Identify ambiguities, conflicts, and missing decisions in a requirement.
3. Map an as-is and to-be process, including handoffs, exceptions, controls, and systems.
4. Build a stakeholder interview guide that distinguishes facts, preferences, and constraints.
5. Convert a business problem into measurable KPIs, definitions, owners, and data caveats.
6. Reconcile two reports with conflicting numbers.
7. Draft user stories with nonfunctional requirements and edge cases.
8. Create a UAT test scenario matrix.
9. Write a concise decision log entry.
10. Summarize a working session with owners, dates, dependencies, and unresolved items.

For every prompt, specify the exact output format it should request. Require the AI to cite which supplied note or data point supports each conclusion and flag rather than resolve unsupported assumptions. Include up to three clarifying questions only when inputs required for that individual task are absent.

Start with a one-paragraph guide explaining when an analyst should use AI for structuring versus when they must validate with the process owner, data owner, security, legal, or compliance team. Do not fabricate system behavior, policies, data definitions, or stakeholder agreement.

Before answering, check that the 10 prompts are materially different and that each produces an analyst artifact, not generic advice.

Create a workshop kit for [INITIATIVE] with [ATTENDEES] using [PRE-READ MATERIAL]. Produce a 60-minute agenda, neutral questions, a decision register, parking-lot categories, and a requirements capture template. For each question, state what decision or ambiguity it resolves. Include prompts that uncover exceptions, approval rules, data ownership, service levels, and audit needs. Do not infer agreement from silence or turn a preference into a requirement. Check that every agenda item has an intended artifact or decision. Ask up to three questions only if a required workshop detail is missing.

Help me reconcile [REPORT A] and [REPORT B] for [METRIC OR BUSINESS QUESTION]. I can provide [DEFINITIONS, QUERIES, AND SAMPLE ROWS]. Produce a comparison plan with a data-dictionary table, likely causes of variance, ordered validation tests, and a finding log template. Test scope, time zone, grain, filters, joins, exclusions, refresh timing, and null treatment before assuming a calculation error. Do not declare either report correct without evidence. Check that each proposed test can distinguish between at least two plausible causes. Ask up to three questions only for missing inputs.

Create a UAT scenario matrix for [FEATURE OR PROCESS] from [REQUIREMENTS AND RULES], for [USER ROLES]. Return a table with scenario ID, preconditions, steps, expected result, test data needed, priority, owner, and evidence to capture. Cover happy paths, permissions, invalid inputs, duplicate handling, integration failures, notifications, reporting effects, and recovery from interruption. Mark requirements that cannot be tested from the supplied material. Do not invent business rules. Check that each acceptance criterion maps to at least one scenario and each high-risk exception has a test. Ask up to three questions only if necessary.