Prompt library
Writing effective prompts ends with one piece of advice: save your best prompts and build a personal toolkit. This page is that toolkit, started for you. Fifteen prompts across the five jobs every business runs on — email, sales, analysis, operations, and hiring — each following the role-context-task-format structure and ready to copy.
How to use this library#
Every prompt has [BRACKETED] fields marking where your specifics go. Copy the prompt with the button in its header, replace every bracket, paste it into your AI tool, and treat the first response as a draft — follow up with “shorter”, “more formal”, or “expand point 2” until it is right.
These prompts work in any major AI tool. If you have not settled on one yet, Tools and models compares the main options.
Email and communication#
Cold outreach email#
For first contact with a prospect who has never heard of you.
You are a B2B sales representative. I sell [PRODUCT OR SERVICE] to
[TARGET CUSTOMER, e.g. "accounting firms with 10-50 employees"]. Write a
cold outreach email to a [DECISION-MAKER TITLE]. Reference [SOMETHING
SPECIFIC ABOUT THEIR BUSINESS]. Keep it under 100 words. End with a
question, not a pitch. Do not use the words "synergy", "solution", or
"reach out".Why it works: the 100-word cap and the closing-question rule force a message that respects the reader’s time instead of pitching at them.
Difficult news to a client#
For delays, price increases, mistakes — anything you would rather not send.
You are an experienced account manager known for handling hard
conversations well. I need to tell [CLIENT NAME], a client of [HOW LONG],
that [THE BAD NEWS, e.g. "their project will be delivered two weeks
late"]. The cause is [REASON]. What I can offer is [WHAT YOU WILL DO
ABOUT IT]. Write the email: the news in the first sentence, no excuses,
one concrete next step, under 150 words.Why it works: “the news in the first sentence” overrides the model’s instinct to bury bad news under three paragraphs of padding.
Meeting recap email#
For turning messy notes into a recap people will actually read.
You are an executive assistant. Here are my raw notes from a meeting
about [TOPIC]: [PASTE NOTES]. Turn them into a recap email to attendees
with three sections: decisions made, action items (each with an owner
and a deadline), and open questions. Bullet points only, no preamble.Why it works: forcing decisions, owners, and deadlines into separate sections means commitments cannot hide in prose.
Sales and follow-up#
Follow-up sequence after no reply#
For prospects who went quiet after you sent a proposal or quote.
You are a B2B salesperson who follows up without being annoying. I sent
[PROSPECT NAME] at [COMPANY] a proposal for [WHAT YOU PROPOSED] on
[DATE] and have heard nothing. Write three follow-up emails to send over
two weeks: first a light nudge, then a new piece of value (an idea,
article, or relevant result), then a polite close-the-loop message.
Each under 75 words. Vary the structure so they do not read as a
template.Why it works: asking for the whole sequence at once produces three distinct messages instead of three versions of “just checking in”.
Discovery call to proposal summary#
For converting call notes into a one-pager the prospect recognizes themselves in.
You are a sales engineer who writes clear one-pagers. Here are my notes
from a discovery call with [PROSPECT NAME] at [COMPANY]: [PASTE NOTES].
Write a one-page summary with: their situation in their own words, the
three problems we can solve, what we propose, and what happens next.
Use their language from the notes wherever possible. No jargon from
our side.Why it works: “use their language from the notes” makes the prospect feel heard, which generic proposal copy never does.
Objection-handling prep#
For preparing before a sales call instead of improvising during it.
You are a sales coach. I sell [PRODUCT OR SERVICE] at [PRICE POINT] to
[TARGET CUSTOMER]. List the eight objections I am most likely to hear,
ordered by how often they come up. For each one, give a two-sentence
response that acknowledges the concern before answering it, and one
follow-up question to keep the conversation going.Why it works: requiring acknowledgement before the answer produces responses that sound like listening, not a script.
Analysis and summarization#
Contract or document risk review#
For a first pass over a contract, lease, or supplier agreement before the professionals see it.
You are a commercial advisor reviewing a document for a small business
owner. Here is the document: [PASTE DOCUMENT OR ATTACH FILE]. Identify
the five clauses or sections that carry the most risk for me as the
[YOUR ROLE, e.g. "buyer"]. For each: quote the relevant text, explain
the risk in plain language, and rate it high, medium, or low. End with
the three questions I should ask before signing. This is preparation
for a professional review, not a replacement for one.Why it works: asking the model to quote the text it flags makes the output checkable — you can verify every claim against the source.
Financial trend analysis#
For making sense of a P&L without waiting for your accountant.
You are a CFO advisor to a [INDUSTRY] business with roughly [ANNUAL
REVENUE] in revenue. Here is my P&L for the last [PERIOD]: [PASTE DATA].
Identify the three line items with the largest change versus the prior
period. Present them in a table with: line item, prior value, current
value, percentage change, and one suggested action each. Flag anything
that looks like a data error rather than a real change.Why it works: “flag anything that looks like a data error” catches the copy-paste mistakes that would otherwise drive the whole analysis.
Customer feedback themes#
For finding the signal in a pile of reviews and survey responses.
You are a customer insights analyst. Here are [NUMBER] customer reviews
and survey responses for my [BUSINESS TYPE]: [PASTE FEEDBACK]. Group
them into themes, ranked by how often each theme appears. For each
theme: a one-line summary, one representative quote, and whether it is
a complaint, a request, or praise. Finish with the single change that
would address the most feedback.Why it works: the ranking and the representative quotes turn a pile of anecdotes into something you can act on and defend.
Ops and SOPs#
Turn a process into an SOP#
For getting the process out of your head and into a document someone else can follow.
You are an operations manager who writes procedures people actually
follow. I am going to describe a process we run at my [COMPANY TYPE]
from memory: [DESCRIBE THE PROCESS, INCLUDING WHO DOES WHAT]. Turn it
into a standard operating procedure with: purpose (one sentence), who
does this and when, prerequisites, numbered steps with one action each,
and a "common mistakes" section. Before writing, ask me up to five
clarifying questions if anything is ambiguous.Why it works: inviting clarifying questions first surfaces the steps you forgot to mention — which are exactly the ones new staff get wrong.
Weekly status report#
For turning scattered updates into a report your team or board can scan in a minute.
You are a chief of staff. Here are my rough notes on what happened this
week at [COMPANY OR TEAM]: [PASTE BULLET UPDATES]. Turn them into a
status report with four sections: wins, in progress, blocked (with what
is needed to unblock), and next week's priorities. Keep each item to
one line. If something I wrote is vague, keep it but mark it with [?]
so I know to firm it up.Why it works: the [?] rule stops the model from inventing specifics to fill the gaps in your notes.
Decision memo#
For decisions big enough to deserve more than a gut call.
You are a strategy consultant who writes tight one-page memos. I need
to decide: [THE DECISION, e.g. "whether to hire a second technician or
outsource overflow work"]. The options are [OPTION A], [OPTION B], and
doing nothing. Context: [RELEVANT FACTS, BUDGET, CONSTRAINTS]. Write a
memo with: the decision in one sentence, a comparison table of the
options (cost, effort, risk, reversibility), your recommendation with
reasoning, and what would change your mind.Why it works: “doing nothing” as an explicit option and “what would change your mind” force honest analysis instead of a justification for the choice you hinted at.
Hiring#
Resumes and interview notes contain personal information. If your company has rules about what can be pasted into AI tools — and it should: see Writing your team’s AI usage policy — check them first, and keep every hiring decision with a human.
Job description#
For a posting that attracts the right people and filters the wrong ones.
You are an HR manager at a [SIZE]-person [INDUSTRY] company. Write a
job description for a [JOB TITLE]. Include: a two-sentence summary of
why this role matters, day-to-day responsibilities, must-have
requirements (no more than six), nice-to-haves, and the salary range
of [SALARY RANGE]. Tone: straightforward and honest. No "rockstar",
no "fast-paced environment", no "wear many hats".Why it works: capping must-haves at six keeps the posting realistic, and banning the cliches keeps it readable.
Interview question bank#
For interviewing with a plan instead of winging it.
You are an experienced hiring manager. I am interviewing candidates for
a [JOB TITLE] role where the most important skills are [SKILL 1],
[SKILL 2], and [SKILL 3]. Write ten interview questions: six
behavioural ("tell me about a time..."), three practical or
scenario-based, and one about how they would approach their first
month. For each question, add one line describing what a strong answer
sounds like.Why it works: “what a strong answer sounds like” turns a question list into a scoring guide, so different interviewers judge consistently.
Candidate screening summary#
For comparing a resume against your requirements before booking a call.
You are a recruitment screener. Here is the role's requirements list:
[PASTE REQUIREMENTS]. Here is a candidate's resume: [PASTE RESUME].
Produce a table matching each requirement to evidence from the resume
— met, partially met, or no evidence — quoting the resume where
possible. Then list three questions to probe the gaps in a screening
call. Do not infer skills that are not stated.Why it works: “do not infer skills that are not stated” stops the model from generously filling gaps the candidate never claimed.
Make these your own#
These prompts are starting points, not scripture. Swap the role, tighten the format, add your company’s context — and when a modified version works, save it. Your edited library will beat this one within a month, because it will be tuned to your business.
If you want to push further, Advanced techniques covers chain-of-thought reasoning and few-shot examples, and Exercises gives you practice tasks to build the habit.