TTaras

A practical guide · Taras Shilo · 16 November 2026

AI prompts for work: practical examples

A useful prompt states the outcome, approved inputs and review criteria. These original templates are starting points: replace the placeholders and work with anonymized examples.

Define the task before the role

State the audience, source material, constraints and desired format. Include a short example when structure matters. An instruction to avoid mistakes does not replace checking facts, figures and sources.

Draft a customer email

Review the recipient, commitments and dates before sending.

A template for your task

Draft a customer email.
Situation: [anonymized context].
Confirmed facts: [list].
Next step: [action].
Use clear, calm language. Do not invent deadlines, discounts or guarantees. List missing facts separately.
Output: a subject and a draft of up to 150 words.

Explore marketing hypotheses

Ask the model to separate evidence from interpretation.

A template for your task

Review these anonymized comments: [text].
Identify recurring problems and quote supporting fragments. Suggest three improvement hypotheses.
For each: observation, hypothesis, test and risk. Do not invent reviews or audience percentages.

Create an onboarding draft

Use organizational instructions rather than personal assessments of employees.

A template for your task

Use this policy [anonymized text] to draft an onboarding plan.
Columns: stage, task, expected outcome, owner. If an owner or date is missing, write “needs agreement”. List contradictions and open questions.

Summarize a report

Check calculations against the source table.

A template for your task

Summarize [anonymized data]. Separate confirmed changes, possible explanations and next checks. Do not treat correlation as causation or fill gaps with invented values. Link every numeric conclusion to a source row or metric.

Test repeatability

Try several typical tasks and one difficult case. Compare factual accuracy and editing effort. A recurring prompt can become a documented team procedure or a skill with examples and checks.

Common questions

Will the same prompt work in every model?

The task can be reused, but outputs depend on the model, tools and context. Test it in the service the team will actually use.

Can I upload customer documents?

Check organizational policies and service terms first. Use anonymized or synthetic examples while learning.

Related documentation

The examples and review approach are original. Check each tool’s documentation for current capabilities.