Looking for the most useful AI prompts for work? Below is a curated library of 100 battle-tested prompt templates across six categories — office productivity, documents and writing, data analysis, content creation, marketing, and finance — each one ready to copy, paste, and adapt in ChatGPT, Claude, Gemini, or any AI assistant.

Here is the uncomfortable truth: the gap between "AI is useless" and "AI is my best employee" is rarely the model you pick. It is the prompt you type. A vague request gets a vague answer; a structured prompt with a role, context, goal, steps, and output format gets a deliverable you can ship. This guide gives you both the formula and the templates.

What Makes a Good AI Prompt? The 5-Part Formula

Almost every high-performing AI prompt follows the same skeleton:

Role + Context + Goal + Steps + Output Format

• Role — assign an expert persona: "You are a data analyst…", "You are a meeting minutes specialist…"
• Context — paste the document, data, or background the AI should work from.
• Goal — state exactly what outcome you want.
• Steps — numbered requirements remove ambiguity and force completeness.
• Output format — Markdown, a table, under 500 words, professional tone. Say it explicitly.

Example — the weekly report prompt, fully assembled:

"You are a workplace reporting expert. Based on [this week's work log], write my weekly report. Requirements: 1. Structure it into four sections: completed, in progress, issues and risks, next week's plan. 2. Quantify key results with data, e.g. 'conversion rate up 20%'. 3. Every issue must include a proposed solution; every risk must include impact level and mitigation. Output format: Markdown, ~800 words, professional and results-oriented."

Best AI Prompts for Office Productivity

The biggest time sinks — reports, minutes, emails, scheduling — are also the easiest to delegate. Three templates that pay for themselves in the first week:

• Meeting minutes: "You are a professional meeting recorder. Turn [transcript] into formal minutes. Extract core conclusions per topic; list action items in a table with owner, deadline, and priority; list disputed points separately with each side's view."
• Email replies: "You are a communication expert. Reply to [this email]: first confirm my understanding of the key points, then respond to each request specifically, then state the next action and timeline. Professional tone, no emotional language."
• Difficult conversations: "You are a workplace communication coach. I need to discuss [sensitive topic] with [person]. Give me 3 scripts — diplomatic, firm, and compromise — each with an opener, core message, and responses to likely pushback."

Also in this set: batch file organization with a preview-and-confirm step, daily time-blocked schedules, project breakdowns with critical path, extracting action items from chat logs, SOP writing, project retrospectives, interview question banks with scorecards, and converting emotional messages into calm professional language.

AI Prompts for Documents and Writing

The go-to summarizer template — a classic for a reason:

"You are a document distillation expert. Summarize [document]: give a one-sentence conclusion, five core points of no more than two sentences each, and three actionable recommendations. Output in Markdown, under 500 words."

The rest of the writing toolkit: contract review that flags unfavorable clauses and missing protections (with the built-in note that it is not legal advice), text polishing with before/after explanations, expanding outlines into reports, compressing long documents while keeping key data, business-style translation with a terminology table plus a second translation-QC prompt to catch mistranslations, PPT outlines with speaker notes, speech scripts at ~200 words per minute, PRDs, user manuals with FAQ, multi-document comparison of consensus and disagreements, and document desensitization with a restore mapping table.

ChatGPT Prompts for Data Analysis

For many analysts, this category is the highest-leverage of all. The SQL prompt:

"You are a SQL expert. Write a query for [requirement] against [table schema]. Include comments explaining each step, suggest optimizations (indexes, partitioning), and propose an alternative if performance could be an issue."

Also included: data-cleaning plans covering missing values, duplicates, outliers, and format standardization; Excel formulas with every parameter explained; diagnosing why a metric moved (internal vs. external vs. data causes, with verification steps); building a North Star metric tree with definitions and formulas; chart-type selection by data relationship; conversion funnel bottleneck analysis; first-click vs. last-click vs. linear attribution with budget recommendations; RFM customer segmentation; forecast method selection; A/B test design with sample size and common traps like Simpson's paradox; profit decomposition into volume, price, cost, and mix; and turning analysis conclusions into a background-conflict-conclusion-action data story for leadership.

AI Prompts for Content Creation and Marketing

Headlines are where AI earns its keep — one idea, ten angles:

"You are a headline expert (the ethical kind). Write 10 titles for [article]. Each must include a number, suspense, a clear benefit, or emotional pull; tag each with its best platform; avoid exaggerated or false promises."

The rest: 20 content ideas per topic sorted into pain points, tutorials, trends, and stories; social media posts in platform-native voice; 60-second short-video scripts with a 3-second hook; livestream selling scripts on a timeline; brand storytelling; quotable one-liners for posts and covers; five different styles of social copy; breaking down viral posts into reusable patterns; 7-day content calendars; and personal brand positioning with a 30-day launch plan. On the marketing side: user personas with touchpoints, quarterly strategies with phases and KPIs, campaign plans with ROI estimates, ad copy variants, landing page audits with prioritized fixes, community and private-domain playbooks, referral program design, slogans, press releases, crisis statements, partnership proposals, and a full-year marketing calendar.

AI Prompts for Finance and Investing

The template our community will appreciate most — investment appraisal:

"You are an investment analyst. Evaluate [project]: calculate payback period, IRR, NPV, and run a sensitivity analysis. Present optimistic, base, and pessimistic scenarios, and flag the key risks."

Also in the finance pack: monthly budgets separating fixed and variable costs; interpreting income statements, balance sheets, and cash flow statements; cost-structure analysis with reduction priorities; 3-month cash flow forecasting with risk warnings; expense report auditing; invoice-to-ledger field extraction; tax compliance checklists; product pricing strategy (penetration, skimming, bundling); accounts receivable aging with collection scripts; expense control policies; and a financial risk register ranked by probability and impact, with monitoring thresholds.

How to Adapt These Prompts to Your Job

Three rules get you 90% of the value:

1. Replace every placeholder — like [document] or [table schema] — with real, specific context. Specificity in, specificity out.
2. Keep the numbered steps. They are the difference between a checklist-complete answer and a lucky one.
3. Always specify the output format, length, and tone. If you care how it looks, say so before the first word is generated.

FAQ: Using AI Prompts at Work

What is the best format for an AI prompt?
The 5-part structure: role, context, goal, steps, output format. It works because it removes the two biggest failure sources — ambiguity and missing context.

Do the same prompts work in ChatGPT, Claude, Gemini, and Copilot?
Yes. Good prompts are structure-based, not model-specific. You may tune tone or length per model, but the skeleton transfers.

How long should a prompt be?
Long enough to be unambiguous — typically 100–300 words for work tasks. A one-line prompt outsources all the thinking to the model; a structured prompt keeps you in control.

How do I get more accurate AI outputs?
Demand numbered requirements, an explicit output format, and ask the AI to list its assumptions or ask clarifying questions before answering. Accuracy is mostly a prompt-design problem.

Key Takeaways

• Use the formula: Role + Context + Goal + Steps + Output Format.
• 100 ready-to-use templates cover six work areas: office, documents, data analysis, content, marketing, finance.
• Numbers, tables, and explicit formats in your prompt produce shippable deliverables, not essays.
• Copy a template, swap the placeholders, iterate once — most tasks take under five minutes.

Save this page as your prompt reference — and if one template saves you an hour this week, that was the point. Which category should we expand next: quant research prompts or trading workflow prompts? Tell us in the comments.