How to Show Employers You Can Use AI With a Practical Work Sample

Published October 8, 2026 · 3 min read

Saying you use AI tells an employer little about the quality of your work. A small sample can show how you choose a task, provide context, review output, and decide what is ready to use.

Choose a problem relevant to the role. You do not need to build a complex agent to demonstrate sound judgment.

Select a task with a visible output

Useful examples include a customer-support reference guide, a research brief, a spreadsheet analysis, a marketing experiment plan, or a documented process.

Make the input manageable and appropriate to share. Use fictional customer questions, public information, or data you have permission to use. Do not upload confidential employer material simply to make the sample appear realistic.

Write the intended outcome in one sentence. For example: “Create a clear reference guide that helps a fictional support team answer five recurring account questions.”

Show the workflow

Explain what you supplied, what the AI assistant produced, and what you changed. Include a few decisions rather than a long transcript.

A fictional support-guide sample might involve these steps:

1. Define the audience and five questions.
2. Draft answers from a supplied fictional policy.
3. Check every answer against that policy.
4. Remove claims the policy does not support.
5. Add an escalation instruction where information is missing.
6. Test the guide with two ambiguous questions.

This demonstrates more than prompting. It shows whether you can recognize when a fluent answer is unsupported.

Include one error you caught

A polished final artifact can hide the most important part of the work. Show a short example of an output you rejected and explain why.

Perhaps the assistant invented a refund deadline, treated an assumption as a fact, or produced a calculation with the wrong denominator. Describe the correction and the check that caught it.

Do not invent a dramatic failure for the story. If you did not observe one, explain the checks you performed and their limits.

Evaluate the result honestly

Define a practical check: factual accuracy against the source, coverage of the intended questions, calculation correctness, readability, or consistency of terminology.

For a small independent exercise, report only what you actually tested. “All five answers matched the supplied fictional policy after revision” is supportable if you performed that check. “Reduced customer-support costs by 30%” is not supportable without real deployment and evidence.

Keep a distinction between your contribution, the tool’s output, and any other person’s work.

Package the sample

Use a short document with the problem, inputs, final artifact, decisions, checks, and limitations. Add the date and tools used if relevant. Make clear that it is an independent practice project.

Be ready to discuss what would change with real customer data, a larger team, or an ongoing process. Do not claim expertise in security, engineering, or model evaluation beyond what you demonstrated.

Try this AI prompt

Your next step

Pick one small task and write the verification checklist before generating anything. The strongest sample shows that you know how to decide whether the output is useful.