02Prompting · Library، Template

A Library of 30 Prompts for Marketing and Work

Thirty reviewed prompt templates for strategy, research, copy, analysis and operations, each with variables, a sample input, what a good output looks like and an evaluation note.

Who it's for
Marketing teams, small business owners and associations who repeat writing and analysis tasks every week.
Level
Intermediate
Time
45 min
Version
1.0 · 21 September 2026

How to use this library

These are thirty templates for tasks that recur in marketing and daily work, grouped into five sets: strategy, research, copy, analysis and operations. Each one is built on the five-part framework (role, context, task, constraints, format), and each has four parts:

  • Variables: what you replace inside the square brackets with your real information.
  • Sample input: a short fictional example showing the kind of information needed.
  • What a good output looks like: what you should get if the prompt is doing its job.
  • Evaluation note: what to check before you rely on the result, and where models usually go wrong on this kind of task.
An honest note: we reviewed these templates carefully for structure, clarity of instruction and safeguards against invention, but they have not been through systematic benchmark testing on specific tools. Results vary with the tool, its version and the quality of your inputs. Treat them as a starting point: try them on your own material, adjust them and keep the version that works for you.

General rules for every template

  • Fill in every variable; a template with an empty variable produces a generic result.
  • Remove names, phone numbers and sensitive data from anything you paste.
  • Every number, name or claim in the result is checked before publishing or sending.
  • Start a new conversation for each template, and improve results with short follow-ups rather than rewriting.
  • The templates are written in Arabic on the Arabic page; if you want the output in another language or a dialect, say so in the constraints.

1. Strategy

1. Value proposition

When to use it: before a campaign or landing page, when you cannot say in one sentence why a customer should choose you. Variables: [product or service], [audience], [alternatives they use now], [proven features].

Value proposition
You are a marketing strategist who writes in plain, specific language.
Context: we offer [product or service] to [audience].
Our audience currently uses [alternatives], and their problem with them is
[problem]. Our proven features only: [features].
Task: propose 3 value propositions, each with a headline (max 10 words),
one explanatory sentence and 3 reasons to believe taken only from the
features above.
Constraints: do not use "best", "first" or any claim the features do not
support.
Format: table: Version | Headline | Explanation | Reasons to believe |
Who it suits best.
Illustrative example

Sample input: Khayt w Ibra, a clothing alterations shop in Tripoli; audience: busy office workers; alternative: the neighbourhood tailor with unpredictable hours; features: pick-up and delivery within 48 hours, prices published upfront, booking by WhatsApp.

What a good output looks like: three genuinely different versions (time, price clarity, ease), such as "Your alterations done in two days, and you know the price before we start", with reasons to believe drawn only from the three features.

Evaluation note: check every "reason to believe": is it in your inputs? Models tend to add "high quality" and "years of experience" on their own.

2. Audience segments and their needs

When to use it: when you are talking to "everyone" and need a practical split. Variables: [project description], [what we actually know about customers], [what we do not know].

Audience segments
You are a practical marketing researcher.
Context: [project description].
What we know about our customers from observation and sales: [information].
What we do not know yet: [gaps].
Task: propose 3 to 4 audience segments. For each: description, core need,
what stops them buying, the best-fit message and the likeliest channel.
Constraints: clearly separate what rests on our information from what is an
assumption that needs checking.
Format: table, then a list of 5 questions to test the assumptions.
Illustrative example

Sample input: Daraj Café near a university; we know the peak is 10 am to 1 pm and most orders are coffee and small pastries; we do not know why evening traffic is low.

What a good output looks like: segments such as "students between lectures" and "remote workers", a column that explicitly marks "assumption" next to anything not from the inputs, and test questions you can actually ask customers.

Evaluation note: useful segments differ in need and behaviour, not just age. If every segment is "young people who value quality", ask again.

3. Turning a vague goal into measurable objectives

When to use it: when the goal is "grow sales" or "grow on social". Variables: [overall goal], [current baseline], [resources and time], [period].

Measurable objectives
You are a planning adviser who helps small teams.
Overall goal: [goal]. Current situation in numbers: [baseline].
Resources available: [team, time, rough budget]. Period: [period].
Task: turn the goal into 3 measurable sub-objectives. For each: the
indicator, current value, target value, how it is measured and a suggested
owner (a role, not a name).
Constraints: keep targets realistic against the baseline and state the
assumption behind every number. Do not assume missing numbers; if a
baseline is missing, write "must be measured first".
Format: table, then the two biggest risks to achieving it.
Illustrative example

Sample input: Lamsa Fitness Centre; goal: more memberships; 120 active members and about 15 new memberships a month; two trainers and a receptionist; three months.

What a good output looks like: objectives such as "raise new monthly memberships from 15 to 20" with a stated assumption, and a renewal objective marked "must be measured first" because the renewal rate was not given.

Evaluation note: reject any target percentage without a written assumption, and watch for objectives that need resources you never mentioned.

4. SWOT grounded in your facts

When to use it: before a planning session, as a draft for discussion rather than a verdict. Variables: [project description], [city or market], [facts].

Grounded SWOT
You are a cautious business analyst.
For the project [description] in [city or market], and based only on the
facts below:
"""
[facts: what we offer, our team, competitors we know, customer feedback]
"""
Task: propose strengths, weaknesses, opportunities and threats, 3 points
each.
Constraints: each point refers to the fact it rests on. No general market
information.
Format: four headings with bullets, then a list: "what we need to know to
improve this analysis".
Illustrative example

Sample input: Waraqa stationery shop; located near three schools; no online shop; a large competitor opened on the next street; parents ask for school lists to be prepared in advance.

What a good output looks like: an opportunity such as "pre-packed school-list bags" linked to the parents' requests, a threat linked to the nearby competitor, and a list of missing facts such as the competitor's prices.

Evaluation note: any point that does not refer to a fact is generic filler. Delete it or ask where it came from.

5. Campaign messaging pillars

When to use it: to align what everyone writing about the campaign says. Variables: [offer], [audience], [common objections], [tone].

Messaging pillars
You are a campaign strategy writer.
Offer: [offer, duration and terms]. Audience: [audience].
Objections we hear: [objections]. Tone: [tone].
Task: propose 3 messaging pillars. For each: the idea in one sentence,
evidence from the offer that supports it, the objection it answers, and
two example short headlines.
Constraints: no promised outcomes and no terms not stated in the offer.
Format: table, then one sentence that unites the three pillars (the core
message).
Illustrative example

Sample input: Nabd dental clinic; a discounted check-up and cleaning offer during October; audience: families; objections: fear of pain, lack of time, distrust of "offers".

What a good output looks like: pillars addressing fear (explaining the steps in advance), time (visit length, if known) and trust (offer terms written clearly), with no medical claims.

Evaluation note: in sensitive sectors (health, finance, education) review every headline with a qualified person. Models sometimes write implied promises such as "a perfect smile".

6. Prioritising channels

When to use it: when the team is small and the channels are many. Variables: [audience], [goal], [available channels], [weekly hours], [what we tried].

Channel priorities
You are a marketing planner for small teams.
Goal: [goal]. Audience: [audience].
Available channels: [channels]. Time available: [hours per week].
What we tried and how it went, in our own observation: [experiments].
Task: rank the channels from highest to lowest priority and suggest how to
split the hours. For each channel: why it ranks there, the best-fit content
type and one indicator to track.
Constraints: suggest two main channels at most, and assume no ad budget
that is not stated.
Format: table, then what to stop or postpone.
Illustrative example

Sample input: Al Zaytouna Bakery; goal: more morning visits; channels: Instagram, WhatsApp, shop signage, neighbourhood groups; 4 hours a week; we tried TikTok and did not keep it up.

What a good output looks like: two channels close to the neighbourhood (for instance WhatsApp and signage) with a sensible split of hours, and an explicit recommendation to postpone any new channel rather than add load.

Evaluation note: if the model proposes every channel "in equal shares", it has not prioritised anything. A good ranking includes what you stop.

2. Research

7. Customer interview guide

When to use it: before talking to customers to understand a buying decision. Variables: [what we want to understand], [customer type], [interview length].

Interview guide
You are a user researcher who runs short interviews.
We want to understand: [research question]. Interviewing: [customer type].
Length: [minutes].
Task: write an interview guide: a short opening, 8 open questions ordered
from general to specific, and two follow-up probes for each main question.
Constraints: no leading questions ("don't you think...") and no yes/no
questions. Focus on what the customer actually did in the past, not what
they "might do".
Format: numbered list, then 3 mistakes the interviewer must avoid.
Illustrative example

Sample input: Khayt w Ibra; we want to understand how people choose a tailor for the first time; new customers from the last month; 15 minutes.

What a good output looks like: questions such as "Tell me about the last time you needed a garment altered: what did you do first?" rather than "Do you like fast service?".

Evaluation note: read each question and ask: does it push the customer toward the answer we want to hear? If so, rephrase it.

8. Turning customer feedback into themes

When to use it: when comments and reviews pile up and you cannot see what repeats. Variables: [comments with names removed], [source].

Feedback themes
You are a careful customer-feedback analyst.
Inputs: [number] comments from [source], names removed:
"""
[comments]
"""
Task: group the comments into recurring themes (positive and negative).
For each theme: a name, the number of comments that mention it, and two
short verbatim quotes as examples.
Constraints: count only from the comments given; quotes are verbatim and
unedited. Put ambiguous comments under "unclassified".
Format: table sorted by count, then 3 suggested actions tied to themes.
Illustrative example

Sample input: 40 comments on Daraj Café's page, such as "Great coffee but the morning wait is long" and "No space to study".

What a good output looks like: themes such as "morning waiting time" and "seating for studying" with approximate counts, and quotes you can find word for word in the inputs.

Evaluation note: models sometimes miscount. Check the count for at least two themes by hand and search for the quotes in the original text.

9. Competitor comparison from material you collect

When to use it: after you have gathered competitors' pages and offers yourself. Variables: [competitor material], [comparison criteria], [our offer].

Competitor comparison
You are an objective competition analyst.
Here is material we collected ourselves on [number] competitors (their page
text and offers):
"""
[material, each competitor under a heading]
"""
And this is our offer: [our offer].
Task: compare them on: [published price, main promise, audience, channels...].
Constraints: use the material only. If information is missing, write "not in
the material". Do not infer service quality from marketing language.
Format: comparison table, then 3 gaps where we could stand out, each with
its evidence.
Illustrative example

Sample input: page texts from three fictional fitness centres in the city, and Lamsa Fitness Centre's offer.

What a good output looks like: a table with many cells reading "not in the material", which is a healthy sign, and gaps such as "nobody publishes women-only class times clearly", with the source noted.

Evaluation note: do not ask the model for "information about competitors" from memory; it may invent prices and services. The material must come from you.

10. A verifiable desk-research plan

When to use it: when you need information about a market or topic and do not know where to start. Variables: [question], [market or area], [how the result will be used].

Desk-research plan
You are a research librarian helping me plan, not answer.
My question: [question]. Scope: [market or area]. I will use the result for:
[use].
Task: do not answer the question. Instead give me:
1) the question broken into 4 or 5 sub-questions;
2) for each, the most likely source type (official body, sector report,
interviews, field observation);
3) suggested search phrases in Arabic and English;
4) how to check the reliability of what I find.
Format: table, then what can only be learnt by asking people directly.
Illustrative example

Sample input: "Is there demand for a weekly subscription for morning bread delivery in our neighbourhood?"; Al Zaytouna Bakery; to decide whether to pilot the service.

What a good output looks like: sub-questions such as household numbers and morning buying habits, with a clear note that the most accurate answer will come from surveying shop customers, not the internet.

Evaluation note: if the model starts giving figures and statistics, stop it; you asked for a plan. Any number it states needs a source you open yourself.

11. A short survey

When to use it: to gauge the views of more customers quickly. Variables: [purpose], [audience], [channel], [maximum time].

Short survey
You are a survey designer who respects people's time.
Purpose: [the decision this survey will inform]. Audience: [audience].
Sent via: [channel]. It must take no longer than [minutes].
Task: design a survey of 6 to 8 questions: the type of each question
(choice, scale, short text) and its options where relevant.
Constraints: one idea per question, no leading questions, no unnecessary
personal data. The last question is open and optional.
Format: numbered list, then a short invitation message (3 sentences).
Illustrative example

Sample input: Darb Al Khair Association; we want to know the best times for awareness workshops; volunteers and parents; via WhatsApp; two minutes.

What a good output looks like: questions about days and times with clear options and one about desired topics, without asking for a phone number or address.

Evaluation note: look for double-barrelled questions such as "Are the time and place suitable?" and split them. Delete any question that would not change your decision.

12. Executive summary of a long document

When to use it: a long report, contract or manual that a busy manager must understand. Variables: [document], [reader], [decision expected of them].

Executive summary
You are a management consultant writing for a reader with two minutes.
Reader: [reader]. The document is below.
Task: summarise at three levels:
(1) one sentence stating the main idea;
(2) 5 bullets with the most important findings, ordered by importance;
(3) the decisions or actions required of the reader.
Constraints: plain, clear language with no padding. Quote numbers exactly,
with the page or section. Add nothing that is not in the document.
"""
[document]
"""
Illustrative example

Sample input: a 15-page internal report on the results of Darb Al Khair's seasonal fundraising campaign; the reader is the board chair.

What a good output looks like: a clear main sentence, bullets that each point to a section of the report, and specific decisions such as "approve extending the campaign by two weeks".

Evaluation note: pick three numbers from the summary and find them in the document. Then give the summary to a colleague who has not read the document and ask: is this enough to decide?

3. Copy

13. A post from three angles

When to use it: for social posts when you want a real choice. Variables: [platform], [topic], [real details], [audience], [call to action].

Three-angle post
You are a content writer for [platform].
Topic: [topic]. Real details: [times, published prices, features].
Audience: [audience]. Call to action: [what we want the reader to do].
Task: write 3 versions of the post, each from a substantially different
angle (for example: a problem the reader lives with, a short story, a
useful fact).
Constraints: [number] words maximum per version. The first line grabs
attention without a worn-out question. No superlatives ("the best",
"unmatched") and no details not listed above.
Format: each version with its angle named, then a one-line suggestion for
an accompanying image.
Illustrative example

Sample input: Instagram; Waraqa stationery shop opens until 9 pm during the two back-to-school weeks; working parents; "send your school list on WhatsApp and we will prepare it".

What a good output looks like: one version opening on parents after work, one short story, one practical tip, all ending with the same call to action and none inventing discounts.

Evaluation note: if all three versions open the same way, the angles are not different. Ask for "different in substance, not wording".

14. Short ad copy in several variants

When to use it: preparing paid-ad copy for testing. Variables: [offer], [audience], [platform], [character limits], [exclusions].

Ad copy
You are a performance copywriter who follows platform policies.
Offer: [offer and terms]. Audience: [audience]. Platform: [platform].
Task: write 5 headlines (up to [number] characters) and 3 primary texts
(up to [number] words). Each headline uses a different driver: saving,
time, trust, curiosity, ease.
Constraints: no unprovable claims, no direct reference to the reader's
sensitive personal attributes (health, weight, financial situation), and
no terms that are not stated.
Format: table: Headline | Driver | Character count, then the primary texts
numbered.
Illustrative example

Sample input: Lamsa Fitness Centre; a free trial week for new members; office workers aged 25 to 40 in the city; Instagram ads.

What a good output looks like: headlines such as "A full free week to try before you decide" with a character count, and no lines like "Struggling with extra weight?".

Evaluation note: count the characters yourself; models often miscount. Check the platform's current policies, as they change.

15. A newsletter email

When to use it: a regular message to a list of customers or supporters. Variables: [sender], [list], [news or content], [main goal].

Newsletter email
You are an email editor who writes briefly and warmly.
Sender: [organisation]. List: [who the subscribers are and why they
joined].
Available content: [3 to 4 news items or tips with details].
The main goal of this email: [one action we want].
Task: write a complete email.
Constraints: 250 words maximum. One main idea at the top, the rest brief.
Only one main button or link.
Format: 3 subject-line options, preview text (preheader), then the body
with short subheadings.
Illustrative example

Sample input: Darb Al Khair Association; past donors and volunteers; news: 300 school bags distributed, volunteer sign-up open for the winter campaign, thanks to volunteers; goal: sign-ups for the winter campaign.

What a good output looks like: an email that opens briefly with thanks and the achievement and leads to a single button, "Sign up to volunteer for the winter campaign", with no numbers beyond those given.

Evaluation note: more than one main call to action can scatter the reader. Check every number against your inputs.

16. Product description for an online shop

When to use it: product pages that keep repeating the same phrases. Variables: [product name], [specifications], [use], [audience], [frequent customer questions].

Product description
You are an e-commerce copywriter who writes for readers and search engines
alike.
Product: [name]. Exact specifications: [size, material, weight, care].
Main use: [use]. Audience: [audience].
Questions customers ask: [questions].
Task: write the product description.
Constraints: add no specification that is not listed. No unproven health
or environmental claims. Short sentences, using words people actually
search with.
Format: an opening paragraph (40 words), 5 bullets for specs and benefits,
then 3 questions and answers.
Illustrative example

Sample input: Khayt w Ibra; "reusable fabric tote bag"; cotton, 40 × 35 cm, washable at 30 degrees; for daily shopping; frequent question: how much weight can it hold?

What a good output looks like: a description that repeats the specs accurately and answers the weight question with "not specified, please check with the shop" if no load rating was given, rather than an invented number.

Evaluation note: compare every number in the description with the specs line by line. This is where invented figures slip in most often.

17. A WhatsApp message in dialect

When to use it: when Modern Standard Arabic is too formal for your audience. Variables: [dialect], [situation], [information], [what the recipient should do].

Dialect message
You are a friendly customer-service agent who writes in [dialect] the way
people actually speak it.
Situation: [situation]. Information that must come across: [information].
What the recipient should do: [action].
Task: write a WhatsApp message.
Constraints: 4 sentences maximum. Natural dialect without exaggeration or
slang. No more than one emoji. Information exactly as given.
Format: two versions: the first friendly, the second slightly more formal.
Illustrative example

Sample input: Lebanese; reminding a customer to collect an altered dress from Khayt w Ibra; ready from Thursday, shop open until 7 pm; ask her to confirm a pick-up time.

What a good output looks like: a message in natural Lebanese Arabic saying the dress is ready from Thursday, the shop is open until 7 pm, and asking when suits her to drop by, with details matching exactly.

Evaluation note: read it aloud. If it sounds like a translation from formal Arabic or mixes dialects, ask again with an example from your real messages.

18. Rewriting in three tones

When to use it: one text that must reach different audiences. Variables: [original text], [the three audiences], [a sample of your voice].

Three tones
Rewrite the text below in three versions:
(a) formal, for [an official body]; (b) friendly, for [a long-standing
client]; (c) very short, for [a WhatsApp message].
Keep the same information in all three. Add no promises that are not in
the original.
An example of my preferred voice: [paste a paragraph of your writing].
After the three versions, list any information you had to drop from the
short version.
"""
[original text]
"""
Illustrative example

Sample input: Nabd Clinic's notice about changed summer opening hours; audiences: a professional syndicate, regular patients, a WhatsApp message.

What a good output looks like: three versions carrying exactly the same hours, and an explicit list of what was dropped from the short one (such as the emergency number) so you can decide whether to put it back.

Evaluation note: compare the hours and numbers across all three. The most common error here is a small detail changing while the tone is "improved".

4. Analysis

19. Reading a monthly performance report

When to use it: when you have numbers and do not know what to say about them. Variables: [table of figures], [the two periods], [what changed in our work].

Reading performance
You are a cautious marketing analyst who separates observation from
explanation.
These are our figures for two equal periods: [period 1] and [period 2]:
"""
[table: Indicator | Period 1 | Period 2]
"""
What changed in our work between the periods: [campaign, price, season...].
Task: (1) calculate the change for each indicator in value and percentage;
(2) state 3 notable observations; (3) for each, two possible explanations
framed as hypotheses; (4) what extra data would decide between them.
Constraints: never present an explanation as fact. Do not infer causation
from timing.
Format: a table for the calculations, then bullets.
Illustrative example

Sample input: Daraj Café; September vs August: app orders from 210 to 260, average ticket from $6.20 to $5.80; changes: university term started and a breakfast menu was added.

What a good output looks like: correct arithmetic (up 50 orders, about 24%), hypotheses such as "students returning" and "the breakfast menu is cheaper", and a suggestion to split sales by item to decide.

Evaluation note: recalculate the percentages with a calculator or spreadsheet. If the model says "because" instead of "possibly because", ask it to rephrase.

20. Diagnosing landing-page copy

When to use it: before launching a page, or when a page is not doing its job. Variables: [page text in order], [audience], [page goal].

Landing-page diagnosis
You are a UX and conversion-copy reviewer, frank and unflattering.
Audience: [audience]. Page goal: [one action].
The page text in the order it appears from top to bottom:
"""
[text, noting where buttons and images sit in brackets]
"""
Task: assess the page on: clarity of the offer on the first screen, proof,
objections, call to action and friction (what slows the visitor down).
Constraints: every note points to a specific sentence or section. No
generic advice.
Format: table: Problem | Location | Why it matters | Suggested fix |
Priority (high/medium/low).
Illustrative example

Sample input: Nabd Clinic's booking page; headline "Complete care for your teeth"; the booking button only at the bottom; no opening hours mentioned.

What a good output looks like: notes such as "the headline does not say what the visitor is booking or where" and "the booking button is not visible on the first screen", marked high priority with a specific fix.

Evaluation note: the model sees the text, not the design or loading speed. Complete the diagnosis by opening the page on a real phone.

21. Classifying customer messages

When to use it: sorting a backlog of enquiries or complaints. Variables: [messages with names removed], [categories], [urgency rule].

Message triage
You are an organised customer-service coordinator.
Classify the messages below into: [price enquiry, appointment, complaint,
suggestion, other]. Mark urgency: urgent if [rule], otherwise normal.
"""
[numbered messages, no names or phone numbers]
"""
Constraints: one category per message. If two fit, choose the closest and
say so. Do not write replies now.
Format: table: Message no. | Category | Urgency | 8-word summary, then the
count per category, then the most repeated question that deserves a
ready-made reply.
Illustrative example

Sample input: 25 messages Al Zaytouna Bakery received over the weekend; urgent if about a paid order that has not arrived.

What a good output looks like: a clear table that raises late-order messages to "urgent" and notes that the question about za'atar bread times came up often enough to deserve a ready-made reply.

Evaluation note: review by hand everything marked "urgent" and a sample of "normal". Misclassifying one urgent message costs more than any time saved.

22. The strict critic

When to use it: before sending an important proposal, plan or text. Variables: [text], [who will read it], [what we want from them].

Strict critic
You are a strict critic representing [reader]; your job is to save us
from embarrassment, not to flatter us.
We want the reader to [desired action].
Review the text below and list the 3 most serious weaknesses, most serious
first, with the reason for each and a specific fix.
Then list two questions the reader will ask that the text does not answer.
Do not rewrite the whole text.
"""
[text]
"""
Illustrative example

Sample input: a partnership proposal from Lamsa Fitness Centre to a local company offering discounted memberships for its staff; the reader is the HR manager; the ask is a meeting.

What a good output looks like: points such as "the proposal does not say what the company pays for" and "it is long before it reaches the ask", and a question such as "what if staff do not use the membership?".

Evaluation note: if all the notes are gentle and shallow, ask again with "assume the reader rejected the proposal; what are the three likeliest reasons?".

23. Comparing two options for a decision

When to use it: when you are torn between two options and need to structure your thinking. Variables: [option A], [option B], [criteria and weights], [constraints].

Two-option comparison
You are a neutral decision adviser.
We are choosing between [option A] and [option B].
Criteria that matter to us and their weights (total 100): [criterion:
weight, ...]. Our constraints: [budget, time, team].
Information available on each option: [information].
Task: score each option on each criterion from 1 to 5 with a short reason,
and calculate the weighted total.
Constraints: if there is not enough information to score a criterion, write
"insufficient" and give no score.
Format: table, then: what could change the result, and the smallest
experiment that would help us decide.
Illustrative example

Sample input: Waraqa stationery shop: launch a simple online shop, or set up WhatsApp ordering first; criteria: cost 30, speed 30, customer convenience 25, operational load 15.

What a good output looks like: a table with reasoned scores and correct arithmetic, a note that "customer convenience" needs asking customers, and a suggestion to trial WhatsApp for a month before deciding.

Evaluation note: recalculate the weighted total yourself. The decision is yours: the table structures the thinking, it does not replace it.

24. Cleaning data and building a spreadsheet formula

When to use it: a messy sheet, or a calculation whose logic you know but whose formula you do not. Variables: [column descriptions], [sample rows], [required result], [software].

Spreadsheet formula
You are a spreadsheet expert who explains clearly to non-specialists.
Software: [Excel or Google Sheets]. Columns: [column A: ..., column B: ...].
A sample of 5 rows (non-sensitive data):
"""
[sample]
"""
The problem: [e.g. dates in mixed formats, duplicate names spelled
differently]. What I need: [e.g. total sales per item per month].
Task: propose cleaning steps, then the formula with an explanation of each
part.
Constraints: state your assumptions about the data and flag cases that
could break the formula.
Format: numbered steps, then the formula, then how to verify it by hand on
two rows.
Illustrative example

Sample input: Al Zaytouna Bakery's sales sheet; the date column contains both "1/9" and "2026-09-01"; the item column has "za'atar" and "za'atar bread"; need a monthly total per item.

What a good output looks like: steps to standardise dates and names first, then an explained conditional-sum formula, and a way to compare its result with a manual total for one item.

Evaluation note: test the formula on a copy of the file, not the original, and compare its output with a manual count on a small sample.

5. Operations

25. From meeting notes to tasks

When to use it: after every meeting, before decisions get lost. Variables: [notes], [role names], [meeting date].

Notes to tasks
You are a meticulous administrative coordinator.
Turn the notes from the [date] meeting below into:
(1) decisions taken;
(2) a task table: Task | Owner | Deadline | Priority (high/medium/low);
(3) open questions.
Constraints: do not invent an owner or a deadline; if not mentioned, write
"not specified". After the table, point out any two tasks that seem to
conflict or are unclear.
"""
[notes]
"""
Illustrative example

Sample input: scattered notes from a Lamsa Fitness Centre team meeting about an open day: "Rami to talk to the printer", "need to decide the time", "posts a week before".

What a good output looks like: a task "contact the printer" owned by "Rami" with deadline "not specified", an open question "date of the open day", and a note that the posts depend on that decision.

Evaluation note: look in the table for any date or name not in the notes. Its presence is a serious error even if it looks logical.

26. Ranking the week's tasks by impact

When to use it: when the list is longer than the time. Variables: [tasks], [available time], [the most important result this week].

Task ranking
You are a practical planning assistant.
My tasks this week: [list, with any known deadlines].
Time actually available: [hours]. The most important result I want this
week: [result].
Task: rank the tasks by impact on that result, then by urgency; estimate
the time for each; and suggest what can be postponed, delegated or dropped.
Constraints: if the total exceeds the available time, say so plainly and
do not squeeze the estimates.
Format: table: Task | Impact | Urgency | Estimated time | Decision
(now/postpone/delegate/drop).
Illustrative example

Sample input: the owner of Khayt w Ibra: 12 tasks across customer orders, fabric ordering, product photos and message replies; 30 hours available; result: every wedding order delivered on time.

What a good output looks like: wedding orders at the top, product photos postponed, and an explicit warning if the total exceeds 30 hours.

Evaluation note: time estimates are suggestions; correct them from experience. The template's value is in the questions it forces you to answer.

27. A standard operating procedure from a spoken explanation

When to use it: documenting a task only one person knows. Variables: [the explanation as spoken or written], [who will carry it out], [tools].

Operating procedure
You are a procedure writer writing for a new employee on day one.
This is our colleague's explanation of how to [task], as they said it:
"""
[explanation]
"""
Carried out by: [role]. Tools used: [tools].
Task: turn the explanation into a procedure: purpose, when it runs, what
is needed before starting, numbered steps (one action per step), how we
know it is complete, and what to do if something goes wrong.
Constraints: add no steps that are not in the explanation. Insert
[needs clarification] where the explanation is incomplete.
Format: short headings and numbered steps, then a list of questions to ask
the colleague.
Illustrative example

Sample input: a recorded explanation from the cashier at Daraj Café on how to close the day and reconcile sales.

What a good output looks like: clear steps such as "print the end-of-day report from the point-of-sale system", with [needs clarification] at points such as "what do we do if the amount does not match?".

Evaluation note: ask someone who does not know the task to carry it out from the procedure alone. Every question they ask marks a missing step.

28. Replying to a customer complaint

When to use it: a public or private complaint that needs a calm, specific reply. Variables: [complaint text], [what actually happened], [what we can offer], [channel].

Complaint reply
You are a calm, fair customer-relations officer.
The complaint as received via [channel]:
"""
[complaint text without personal data]
"""
What actually happened, per our information: [facts].
What we can offer: [possible remedy].
Task: write a reply that acknowledges the problem, explains without long
justification, and offers the remedy and its next step.
Constraints: 5 sentences maximum. Do not blame the customer. Promise
nothing not listed above. If the channel is public, move details to a
private conversation.
Format: the reply, then two points we must fix internally so it does not
happen again.
Illustrative example

Sample input: an angry comment on Al Zaytouna Bakery's page: a birthday cake order arrived an hour late; cause: the delivery time was logged wrongly; remedy: refund of the delivery fee and a voucher for the next order.

What a good output looks like: a reply that apologises for the delay without excuses, states the remedy exactly, invites a private message, and two internal points such as "confirm delivery time with the customer by phone".

Evaluation note: do not publish the reply until someone with authority over the remedy approves it, and make sure the tone does not sound automated.

29. A handoff brief for a colleague or freelancer

When to use it: when you hand work to someone else and want it back as expected. Variables: [task], [background], [materials], [deadline], [acceptance criteria].

Handoff brief
You are a project manager writing instructions that leave no room for
interpretation.
The task I am handing over: [task]. Recipient: [their role and experience].
Background that matters: [what they must know].
Materials available: [files and links]. Deadline: [date].
What they must not do: [exclusions].
Task: write a one-page handoff brief: goal, exactly what is required, what
is not required, materials, deadline and review points, acceptance
criteria, and who to ask when stuck.
Constraints: direct language, no assumptions about working methods I have
not stated.
Format: short headings with bullets, then 3 questions the recipient might
ask that I have not answered.
Illustrative example

Sample input: Darb Al Khair Association hands a volunteer designer the winter-campaign poster; logo and colours available; due in a week; no photos of child beneficiaries.

What a good output looks like: a brief that states sizes if given, repeats the ban on beneficiary photos clearly, and sets acceptance criteria such as "Arabic text readable from two metres".

Evaluation note: the three questions at the end are the most important part; answer them and add them to the brief before sending.

30. An FAQ from an internal document

When to use it: turning a policy or manual into quick answers for the team or customers. Variables: [document], [audience], [number of questions].

FAQ builder
You are a support-content writer who writes short, accurate answers.
Audience: [new staff or customers]. The reference document is below.
Task: extract [number] questions the audience is likely to ask, and answer
each in two or three sentences, naming the section the answer came from.
Constraints: answers from the document only. If the document does not
answer an important question, put it in a separate list titled "Questions
the document does not answer".
Format: numbered Q and A, then the separate list.
"""
[document]
"""
Illustrative example

Sample input: Lamsa Fitness Centre's cancellation and refund policy; audience: new members; 10 questions.

What a good output looks like: questions such as "Can I freeze my membership while travelling?" with an answer taken from a named section, and a separate list of what the policy does not cover, such as long illness.

Evaluation note: the separate list is often worth more than the answers: it shows the gaps in your policy. Review it with the decision-maker before publishing.

Building your own library from these templates

  1. Pick five tasks you repeat every week and find the closest template for each.
  2. Fill in the fixed variables (organisation name, audience, tone) once and save them.
  3. Try each template on real material three times and note what you changed each time.
  4. Fold recurring edits into the template itself so it needs less adjusting.
  5. Save an example of a good result with each template; the example lifts the quality of the next use.

General criteria for judging any result

CriterionCheck question
AccuracyIs every number, name and claim in the inputs or verified?
FitDoes the result suit the stated audience and goal?
ComplianceDid it respect the length, format and exclusions?
PrivacyIs it free of personal or confidential data that does not belong there?
UsabilityCan it be used with light editing, or does it need rewriting?

Common mistakes

  • Mistake: using a template with empty or generic variables. Fix: fill every variable with real, specific information, or delete the line.
  • Mistake: asking the model for market or competitor information from memory. Fix: collect the material yourself, paste it, and ask for "not available" instead of guesses.
  • Mistake: trusting counts and calculations. Fix: redo every calculation in a spreadsheet or calculator and check counts on a sample.
  • Mistake: publishing ad copy without checking platform or sector rules. Fix: check the current policies, and have copy in sensitive sectors reviewed by a qualified person.
  • Mistake: pasting real customer data into a public tool. Fix: remove names, numbers and identifying details before pasting.

Checklist before relying on any result

  • I filled in every variable with real information.
  • I removed sensitive data from the inputs.
  • I checked every number, name and claim in the result.
  • I redid every important calculation or count myself.
  • I confirmed the result respected the constraints and format.
  • A person read the result before it was published or sent.
  • I saved the adjusted template that worked, with an example of its output.

Next step

Want a view on your own situation? Marketing Strategy — a 60-minute session.

Book a strategy session

Free to use in your work and organisation; credit the source if you republish.