Proof You Can Explain

Your best sample of work may be smaller than you think.

It does not have to be a complete business, a polished app or a report that takes a stranger an afternoon to read. It needs to give the right person a fair view of something you can do. A small problem, handled carefully and explained honestly, can provide the start of that view.

I say start because a sample is limited evidence. It may show how you handled one task with plenty of time. It cannot establish how you will manage every part of a job. A glossy presentation might reflect purchased help, an excellent template or the work of an AI system. Even a wholly independent sample may omit the hardest part of employment: doing useful work with other people under real constraints.

The answer is to make the sample inspectable. Let a reader see enough of the process to ask a worthwhile question.

This chapter offers one way to do that. It is a practical proposal, not a hiring method whose success has been established by an experiment. Use it where a work example fits the occupation. Some roles require formal assessment, supervised practice or a license. A folder on the internet cannot replace those requirements.

Choose a problem with edges

Suppose you want an entry role that involves basic records, customer communication and office software. You decide to make a booking sheet for a small community room.

Keep it invented. Create a short set of sample bookings rather than copying names, contact details or records from a real organization. The sample needs enough variety to raise a few decisions: two bookings that overlap, one request without an end time, a cancellation and a repeated weekly booking.

Your purpose is to help a coordinator identify which requests can be accepted and which need follow-up. That sentence sets a useful boundary. You are not building the whole organization. You are not claiming to solve payment security, building access, legal duties or every possible calendar problem. You are showing how you turn a modest brief into something a person could inspect.

Before using software, write what a successful result should let the coordinator do. They should be able to spot a conflict, see missing information and avoid treating an unconfirmed request as a confirmed booking. Choose a few requirements you can actually test.

This preparation can feel less impressive than starting the design. It is also where a lot of the judgment becomes visible. If you cannot say what the thing should help someone do, you will struggle to explain why your version is good.

A person entering a different field should choose a different problem. A design applicant might revise a public-domain information sheet for a clearly named audience. A trainee technician might document a safe practice exercise within approved training limits. Someone seeking a customer-service role might compare two possible replies to an invented complaint and explain when a supervisor should take over.

The task should resemble a piece of the desired work closely enough to be relevant. It should be small enough to finish without donating weeks of unpaid labor to an uncertain audience.

Show where the work came from

The first page of the room-booking example can be quite plain. It describes the task, states that all records are invented and names the tools used. It says which parts you made and what assistance you received.

If AI helped write a formula, say so. Add how you checked it. If a friend explained a spreadsheet feature, that is different from the friend building the sheet. If you used a licensed template, identify it and respect its terms. You do not need to publish every prompt or private conversation, but you should not give a false account of who did the work.

There is a practical reason for this honesty. An interviewer may ask you to alter the example. If you claimed more independence than you have, the conversation becomes a test of whether the claim will survive. If you described the assistance accurately, you can discuss what you learned from it and what you still need to practice.

Tools are part of work. Concealing them is a poor way to demonstrate that you can use them responsibly.

Be equally careful with the source material. Publicly visible does not always mean free to reuse. An organization may have posted a report without granting permission to republish its images or large passages. Someone may have shared a dataset that still contains personal information. Use material whose permission and privacy conditions you understand, or create a clearly labeled sample of your own.

Do not turn a former employer’s private documents into a portfolio. Removing a logo may leave the substance confidential. If showing real work requires permission, obtain that permission before sharing it. Otherwise describe the task at a safe level or make a fresh example that contains none of the private material.

A hiring sample should demonstrate good judgment about information, not begin with a breach of trust.

Leave one correction in view

You build the sheet. At first, the overlapping bookings appear in different colors, but the missing end time creates no warning. You notice this when you test the sample. You add a clear pending status and a note that a request cannot be confirmed until the duration is known.

Keep a brief record of that change.

There is no need to stage a dramatic failure. The point is simply to show that the first version met a limit, you recognized it, and the revision addressed it. A reader can then ask why you chose that repair and what else might still go wrong.

This is often more informative than a claim that you pay attention to detail. The detail is right there. It has a consequence. Your response can be discussed.

For each important check, state the expected result. Two confirmed bookings should not occupy the same room at the same time. A request missing an end time should remain pending. A cancellation should release the slot without deleting the record of what happened. Run those checks, record what you observed, and correct problems you find.

If a feature still fails, mark it clearly. An honest limitation is better than a hidden one. You can explain what a fuller version would need while keeping the sample within its original boundary.

Try showing the sheet to someone who is willing to spend a few minutes on it. Ask them to find the pending booking without your help. If they cannot, resist explaining the design immediately. Their difficulty is information about the design. You may need a clearer label rather than a longer defense.

You are testing a small claim: can this person find what the sheet is supposed to make easy to find? That is narrower, and more useful, than asking whether they like your work.

Prepare to change something

Now imagine that an interviewer asks what happens if the room closes early on Wednesdays.

You do not need to have anticipated every possible change. You do need to reason about this one. You might explain that the opening hours should be held separately from individual bookings, so a change can be checked against them. You might identify existing bookings that need review before altering the live schedule. You might admit that your first version has no automatic check for this and show how you would investigate one.

The quality of the conversation depends partly on the interviewer. A fair interviewer gives enough context, allows relevant questions and assesses the work against a stated need. They should not reward whoever guesses an unstated preference fastest.

For the applicant, the preparation is to understand the choices already made. Why did you keep a cancelled record? Why did you label the missing time as pending? What happens if a user sorts one column without the rest of the sheet? Which parts need protection or a clearer instruction?

You can practice explaining these choices without turning the explanation into a memorized sales pitch. Ask someone to interrupt with a genuine question. If you lose your place, return to the problem the sheet is meant to solve.

A useful explanation has room for uncertainty. I do not know whether that feature is available, but I would check the documentation and test it on a copy, is a responsible answer when it is true. It is stronger than inventing a feature to keep the conversation moving.

There will still be things you cannot show in a short sample. Reliability over time, care under pressure and cooperation with colleagues need other evidence. The sample earns a closer look. It should not be made to carry the whole hiring decision.

A better sample cannot create more openings

Indeed reports opposite year-over-year trends in U.S. entry and senior postings. Advertisements are not hires, and this comparison does not establish AI causation.¹

Postings diverged by seniority

Measure	Value or meaning
Market	US Indeed postings, across occupations
Baseline	Each group: May 2025 = 100
Comparison	May 2026; common index scale 0–150
Senior	114.7
Entry	92.5
Classification	Algorithmic categories using tasks, responsibilities and qualifications.
Limits	Indices show relative change, not group sizes or hires. AI effects are not isolated.
Baseline mark	Dashed mark: baseline 100
Axis ticks	0, 75, 150

Source: Indeed analysis

Figure 2. Entry and senior postings moved in opposite directions.

This is why advice about proof needs a limit. You can improve how clearly you present your ability while the pool of available opportunities moves against you. An unsuccessful application cannot, by itself, tell you which of those forces mattered.

Picture two applicants with equally careful examples, applying to a team that can afford one new employee. At least one will be rejected. No amount of positive thinking removes the arithmetic. The rejected applicant may still have learned something worth using elsewhere. They should not be told that the outcome proves they failed to want the job enough.

Track what you can observe. Did anyone inspect the sample? Did the response point to a specific missing skill? Did an employer offer a paid next step? Did the advertisement disappear, or remain open without a reply? These clues will not reveal every motive. They can help you decide whether to improve the work, change the target or stop spending time on a route that provides no useful signal.

Set a preparation limit before beginning. The right limit depends on your circumstances and the task, so I will not prescribe a magic number of hours. Decide what a finished first attempt looks like. Once it meets that standard, use it. Repeated polishing can become a way to avoid a discouraging market while consuming time you cannot easily spare.

You need evidence from the outside as well as improvements on the inside.

When the employer wants a sample

A reasonable hiring exercise should be relevant, bounded and clear about how it will be assessed. If it asks for substantial effort or produces work the organization can use, the employer should pay for that work. My recommendation is that employers budget for meaningful trials rather than make unpaid production a hidden condition of access.

An applicant can ask what the exercise is intended to test, how long it should take, whether tools are allowed, who will see it and whether it will be used commercially. Ask about compensation before accepting a large assignment. A genuine assessment does not need to hide these basic terms.

Consider the difference between an invented booking problem and a request to rebuild a company’s live scheduling system. One can test a relevant skill in a short, protected setting. The other may be a piece of productive work. Calling both a challenge does not make their costs the same.

The employer should also provide a way to request an accessible format or a reasonable scheduling adjustment. A timed screen exercise may test something useful in one role and mostly test comfort with that format in another. The assessment should match what the job actually demands.

There are limits to what an applicant can negotiate alone. Someone who needs work urgently may accept terms they would reject from a stronger position. This is one reason standards belong at the employer and program level. Fairness should not depend on each beginner winning a difficult negotiation before the first interview.

If an exercise seems excessive, you can propose a smaller alternative: discuss an existing sample, complete a narrower task or take a paid trial with explicit boundaries. The employer may decline. You have still made the cost visible, which helps you choose what to do next.

Make it easy to inspect, then let it go

The finished room-booking example needs a short entrance. State the problem and give the reader a way to see the result. Put the checks and a before-and-after change nearby. Offer a concise explanation of assistance, remaining limits and the next feature you would test if the job required it.

Avoid attaching a great mass of files simply because producing them took effort. The reader does not know which one matters. Helping them find the important part is itself a piece of communication work.

Keep a private copy of your process where appropriate, but publish only what is safe and useful. A portfolio does not need to expose your home address, the details of a difficult personal history or the private information of someone who helped you. You are presenting evidence of work, not applying for permission to have a life.

When you send the example, connect it to the role in a few accurate sentences. Explain which small part of the work it resembles. Do not call yourself an expert because the file looks finished. Do not apologize for every limit either. A beginner is allowed to have a beginning.

The strongest conversation may concern the part you changed. Someone asks why you noticed the missing end time. You show the test, explain the consequence and discuss what you would do with a more complex booking. At that point, the sample has done something useful. It has moved the conversation from adjectives on a résumé to decisions that another person can examine.

What should follow is a fair chance to do more, with pay and guidance suited to the work. Your sample is an invitation to that chance. It is not a promise to keep working for free until someone is satisfied.

Sources and limits