Which advanced prompts help HR, finance, sales and project managers?

A basic prompt asks for a thing: "Write interview questions." An advanced prompt describes a working situation. A Beginner's Guide to ChatGPT builds its advanced scenarios this way: a role, the context, the problem with a first generic answer, and a refined request that fixes it. The prompts below follow that pattern and add two things that matter at work: placeholders in square brackets, so one prompt can serve many jobs, and limits that stop ChatGPT from inventing what it doesn't know.

Each prompt has the same five parts:

HR: a structured interview guide

Generic interview questions are easy to find. What saves time is a guide built around the actual role, with a way to score answers fairly.

You are an experienced HR manager who designs structured interviews.
Context: we are hiring a [JOB TITLE] at [COMPANY TYPE AND SIZE]. The three must-have
skills are [SKILL 1], [SKILL 2] and [SKILL 3]. Our values are [VALUES].
Deliverable: an interview guide with two behavioural questions per skill
("Tell me about a time when…"), and for each question, what a strong answer includes.
Limits: no questions about age, family, health, religion or other personal matters.
Format: a table with columns Skill, Question, Signs of a strong answer, Score 1–5.

Finance: explaining a budget variance

A variance is the gap between what was budgeted and what was actually spent or earned. Writing the commentary for a monthly report is a good job for ChatGPT, as long as it works only from your numbers.

You are a financial analyst preparing a monthly management report.
Context: here is our budget versus actual for [MONTH], by department, with names removed:
[PASTE TABLE]
Deliverable: (1) the three largest variances, in dollars and percent; (2) possible
explanations for each, marked as possibilities to confirm; (3) one question to ask each
department head; (4) a summary for executives.
Limits: use only figures from the table. If a number you need is missing, say so.
Format: numbered sections; the summary under 100 words.

Check every calculation against your spreadsheet. ChatGPT can slip on arithmetic, and the "possible explanations" are hypotheses, not findings.

Sales: preparing a discovery call

A discovery call is the first real conversation with a prospect, where you learn their needs before you pitch.

You are a senior B2B sales coach.
Context: I sell [PRODUCT] to [CUSTOMER TYPE] in [REGION]. Our main strength is [STRENGTH].
I have a first call with [PROSPECT ROLE] at a [COMPANY SIZE] company next week.
Deliverable: eight open discovery questions, the three objections I'm most likely to
hear with a short honest reply to each, and a follow-up email I can send after the call.
Limits: no claims about competitors; no promises on price or delivery dates.
Format: three headed sections; the email under 120 words with a friendly, direct tone.

Project management: a risk register

A risk register is a list of what could go wrong on a project, how likely and serious each risk is, and what you will do about it.

You are a certified project manager.
Context: we are running [PROJECT] with a team of [NUMBER] and a deadline of [DATE].
Key milestones: [LIST]. Known constraints: [BUDGET, STAFF, SUPPLIERS].
Deliverable: a risk register with the ten most relevant risks, then a three-line status
update for the project sponsor.
Limits: assign owners by role, not by name. Flag any risk that depends on information
I haven't given you.
Format: a table with columns Risk, Likelihood (L/M/H), Impact (L/M/H), Mitigation, Owner role.

Before you paste: employee and client data

Work prompts tempt people to paste real records. Don't, unless your company has approved a tool for it. Leave out names, salaries, performance notes, health information and client details; use roles, ranges and placeholders instead.

Know your settings too. Since April, ChatGPT lets you turn off chat history: those conversations are not used to train OpenAI's models and are deleted after 30 days, though they may be reviewed for abuse before then. ChatGPT Enterprise, launched in August, does not train on business data at all and adds company controls such as single sign-on. Neither setting replaces your own employer's policy on confidential information.

How do you adapt them to your company?

Start by writing a context block: a short paragraph about your organization that you paste above any prompt. Include what the company does, its size and region, the tone it uses with customers, key terms people use internally, and any rules that always apply, such as "Canadian spelling" or "never quote prices." The book's advice to give background about your role, industry and task is exactly this, written once and reused.

Next, replace generic brackets with your real categories. If your company scores interviews on a 1-to-4 scale, or ranks risks as red, amber and green, put that in the format line. ChatGPT tends to follow a format you show more closely than one you only describe.

Then test before you trust. Run each prompt on two or three past cases where you know the right answer, and adjust the limits wherever it went wrong. Keep the working versions in a shared document so colleagues use the same prompts, and update them when your processes change. A prompt library that reflects how your team really works is worth more than any list of clever prompts found online.