Ready-Made Prompts for Business: Sales, Analysis, Planning

Which prompt is chosen for which business job?

OpenAI's current prompting guide recommends splitting complex work into smaller, focused requests and writing the context and desired output clearly. Business work adds one more part: who approves the answer, and what is the consequence of an error?

5 prompts for sales and customer research

1. Discovery call plan

   "Build a 25-minute discovery call plan from this customer brief: [brief]. Give questions first about today's situation, then the problem, the impact, prior attempts, the decision process and the timeline. Don't add a pitch or pricing section. List the information missing from the brief separately."

2. Customer interview

   "Write 12 non-leading interview questions for [product]. Instead of future intent like 'Would you use it?', ask when they last had the problem, how they solved it, how much time/money they lost and why they chose the alternative."

3. Sales objection analysis

   "Group the last 30 days' objections by topic: [anonymised notes]. For each group write the frequency, the real phrase the customer used, the clarifying question behind it and the proof required. Don't assume the objection's cause beyond the text."

4. Post-call decision note

   "Extract only the confirmed information from this call note: [note]. Fill in the fields: situation, goal, obstacle, decision-makers, what was said about budget, date and next step. Write unknown in any field not mentioned."

5. Proposal skeleton

   "Build a proposal document structure for [customer problem]. Give sections for situation, goals, scope, deliverables, milestones, acceptance criteria, inputs required from the customer, out-of-scope work and commercial terms. Don't invent guaranteed outcomes."

5 prompts for analysis and decision preparation

6. Weekly pipeline audit

   "Check this sales table by stage, amount, last activity date and next step: [table]. Separate the opportunities inactive for over 14 days, those with no next step and those past their close date. Don't change the forecast on your own."

7. Diagnosing a KPI change

   "The metric [KPI] changed by [change] during [period]. Verify the calculation, split the change by channel/product/segment and write three possible drivers. For each driver show the evidence for it, the evidence against it and the extra data needed. Don't present correlation as causation."

8. Thematic analysis of customer feedback

   "Group these anonymised reviews by topic, impact and product stage: [reviews]. Add the frequency, three short quotes and a dissenting view to each topic. Don't claim a few loud reviews represent all customers."

9. Decision matrix for alternatives

   "Compare [alternatives] in a weighted table against [criteria]. Show the weights and the source of each score first. Don't invent a score for a cell without evidence; write no data. Before the final decision, check at which weight change the outcome flips."

10. Pricing and scenario analysis

    "Build base, low and high scenarios for [product]. Inputs: [customer count, price, churn, costs]. In each scenario show the assumptions, the calculation formula and the two most sensitive variables. Label this result as a decision test, not a financial forecast."

5 prompts for planning and process

11. 7-day pilot

    "Write a 7-day pilot plan for [hypothesis]. Show today's baseline metric, the test group, the one factor being changed, the acceptance threshold, the stop condition, the responsible person and the daily log format."

12. 90-day priority plan

    "Build a 90-day plan for [goal]. Split the work into three outcome-based phases; add an owner, deliverables, dependencies, metrics and a 'won't do' list to each phase. Don't assign the same resource to two parallel priorities."

13. Process map

    "Map [process] as trigger, input, step, decision, approval, output and exceptions. At each stage write the system used and the responsible role. Flag steps with no known owner or rule separately."

14. Risk register

    "Prepare a risk register for [project]: event, cause, impact, early signal, likelihood, severity, owner and response step. If there's no evidence for a number, don't invent a precise percentage; write a qualitative level and the reasoning."

15. Pre-meeting decision pack

    "Write a one-page meeting pack for [decision]. Include the problem, the current situation, three options, each option's evidence and risk, the rollback path, a recommendation and the question the manager must answer. Keep information sharing separate from the decision."

How is a business prompt tested before use?

1. Pick one real task

   Take work the team actually did last week, not a comfortable demo example. Keep the input material and the time the old method took.

2. Write the acceptance criteria before the answer

   Define factual accuracy, completeness, format, editing time and the critical-error threshold. Adjusting the criteria after seeing the result breaks the test.

3. Feed typical, incomplete and risky inputs

   The normal case alone is not enough. See where the system stops on samples with missing dates, contradictions and sensitive information.

4. Count human editing separately

   Record how many substantive fixes and how many minutes it took to reach a usable version — not just the first answer.

5. Keep the prompt versioned

   Write the date, model, prompt, test samples and decision together. When the model updates, retest critical work with the same test set.

To improve the prompt's structure, use the 10 rules of prompt writing; for a broader example library, the 50 ready-made ChatGPT prompts.

How are privacy, wrong decisions and human approval protected?

Don't upload customer lists, employee performance data, bank statements, contracts or access keys to a personal chatbot account. Minimise and anonymise the data first. Then check the storage and training terms of the account type you use.

OpenAI's Enterprise Privacy documentation states that business products and API data are not used for model training by default. Personal accounts are governed differently; the Data Controls FAQ explains the conversation-training option and the Temporary Chat rules. Contracts and internal policy can override this general information.

In NIST's AI Risk Management Framework approach, risk management starts from context, measurement and continuous oversight. In a business prompt this turns into simple sentences: which error is critical, who will see it, when is human approval required, and how will the work be rolled back?

Output about pricing, investment, legal matters and employees can be decision support, not final professional advice. In those areas, review by a responsible expert must be kept.