Work Sample Automation: How to Request, Collect, and Pay for Candidate Work Samples

A work sample is something a candidate made that shows how they do the job: code, a design file, an article, a financial model, a campaign plan. Work sample automation means using software to handle the repetitive parts of getting one: sending the request, tracking deadlines, collecting files, recording the candidate's explanation, and paying the candidate.

Key points

• Work samples are among the best single predictors of job performance, and candidates rate them as one of the fairest hiring steps.
• Many roles (sales, operations, finance, product management) rarely have public work samples. Even in fields that do, most real work is private.
• A work sample built for your role is easier to compare across candidates than whatever each person happens to have available.
• A work demo (the finished work plus a recorded walkthrough of how it was made) is harder to fake and shows reasoning, not just output.
• Paying candidates and collecting work through a dedicated platform reduces drop-off and admin work.

Why are work samples a strong hiring signal?

Work samples measure the job directly. A resume describes past work, an interview asks the candidate to talk about work, and a work sample shows the work itself.

Scott Woody, former Director of Engineering at Dropbox, told the Holloway Guide to Technical Recruiting that take-homes are "the truest signal" a company can get, with arguably the lowest false negative rate of any interview format. The guide gives three reasons:

Take-homes give the candidate enough time to do the work in an environment that they're comfortable with, so you eliminate the noise of a whiteboard interview.

It's very hard to hide weaknesses in coding ability in a take-home, and follow-up conversations can tell you almost everything you need to know about how a candidate thinks through problems.

Follow-ups also allow you to weed out and correct for any negative signal or false signal - for example, if a candidate cheats, talking through their work will help you figure that out.

None of that reasoning is specific to engineering. Time, a familiar environment, and a conversation about the decisions behind the work apply to a campaign plan or a financial model as much as to code.

The selection research points the same way. Schmidt and Hunter's 1998 meta-analysis, which covered 85 years of research, found work sample tests to have the highest validity of any single method they studied, at .54. In 2022, Sackett and colleagues showed that a common statistical correction had inflated the figures for every method and revised them all downward. What survived the revision is the pattern that matters for hiring managers: the methods built around what a specific job requires, including work samples, job knowledge tests, and structured interviews, held up best, while unstructured interviews, years of experience, and education credentials sat near the bottom. Those last three are the signals most hiring processes lean on first.

Candidates also prefer work samples. A meta-analysis of 38 samples across 17 countries found that work samples and interviews were the most preferred selection methods among applicants. Applicants tend to see these methods as the most relevant to the job.

Which professions have work samples, and which do not?

Some professions produce visible artifacts as a normal part of the job. Others produce work that is spread across meetings, private systems, and team efforts.

For roles in the bottom half of this table, hiring managers usually have no way to see a candidate's work unless they ask the candidate to create something new.

Public vs. private work samples

Public work samples can be found without asking the candidate. Common sources include:

• GitHub and GitLab for code. GitHub reports more than 180 million developers on the platform.
• Medium, Substack, and personal blogs for writing.
• Behance, Dribbble, and personal portfolio sites for design.
• Kaggle and public notebooks for data work.
• Bylines on news and trade publications for journalists.

Private work samples must be requested and often cannot be shared at all. Examples include client work under NDA, internal documents, private code repositories, financial models, and sales materials.

Even in software, where public work is common, most work is private. GitHub's 2025 data showed that 81.5% of contributions happened in private repositories. Designers face a similar limit: guidance for UX designers recommends password-protecting NDA case studies and sharing access only with trusted parties such as potential employers.

Existing work samples also have limits even when you can see them:

• They vary by candidate. One person shows a large team project, another shows a solo side project. They are hard to compare.
• Authorship is unclear. Team projects rarely show who did what.
• AI makes artifacts less reliable. In Gartner surveys, 4 in 10 candidates said they used AI during the application process, mainly to write resumes, cover letters, writing samples, or assessment answers. Gartner also found that 6% of candidates admitted to interview fraud, and it predicts that one in four candidate profiles worldwide will be fake by 2028.

How to get candidates to create a work sample for your role

A work sample made for your role solves the comparison problem. Every candidate responds to the same brief, so you can review them against the same criteria.

Benefits of a tailored work sample

• Direct relevance. The task reflects what the person will actually do in the role.
• Fair comparison. Every candidate works from the same instructions and constraints.
• Coverage for roles without portfolios. A sales rep, PM, or finance analyst can produce a sample even if their past work is confidential.
• Better candidate experience. Candidates see what the job involves before accepting it.

Make it useful to the candidate, not to your business

The best brief is realistic, gives the candidate something they would be glad to keep, and is not directly useful to your company. Asking candidates to solve a live business problem creates legal and ethical issues and discourages strong applicants. One recruiting expert quoted by InHerSight advises that using a data set the company could use internally is unethical, and suggests using data from a different industry instead.

A good brief meets three tests:

• It is realistic. It uses real tools, real public data, and real companies or products, so the work reflects actual job conditions.
• It is an asset for the candidate. The result is something they could add to a portfolio, publish on GitHub or a blog, or show in future interviews, whether or not they get the job.
• It is not usable by your company. It does not touch your product, roadmap, customers, or internal data. It works on a neighboring problem, a different company, or an adjacent part of your industry.

Also state in writing that the candidate keeps ownership of the work and that you will not use it.

Candidates put more effort into work they can reuse, and a brief that visibly avoids your own business removes the most common concern about take-homes: that the company is getting free work.

Example briefs:

How to get candidates to say yes

Candidates are more likely to complete a work sample when the request respects their time. In a Monster survey, 59% of workers said unpaid assignments or excessive take-home work would make them least likely to apply. However, 66% said mandatory assignments would not necessarily stop them from applying, which Monster read as a sign that candidates will complete assessments that feel fair.

To make the request fair:

• Limit the time. Basecamp has used programming assignments estimated at 3-5 hours and asks only 1-3% of applicants to complete them.
• Pay for the work, especially for longer tasks (see below).
• Send it late in the process, only to candidates with a real chance of an offer.
• Explain how it will be reviewed and what "good" looks like.
• Let candidates use their normal tools instead of an unfamiliar test environment.
• Let candidates choose their deadline within a reasonable window.
• Give feedback to everyone who completes it.

How to ask for and collect work samples

Asking for existing work samples

When a candidate has past work they can share, ask directly and give them an easy way out:

"As part of our process, we'd like to see an example of your work that is similar to what this role involves. Please share one or two samples along with a short note on your role in each. If your past work is under NDA, let us know and we'll send a short paid exercise instead."

Common collection methods

Email and shared drives work for one or two hires. Beyond that, the admin load grows quickly: sending briefs, chasing late submissions, renaming files, anonymizing work for reviewers, and copying results back into your ATS. Ashby's 2026 benchmarks found take-homes in about 13% of hires, and they were more common in roles with more applications per hire, which are the roles where manual collection is hardest to manage.

A dedicated platform handles those steps automatically and keeps every submission in the same format, which makes review faster and more consistent.

Work samples vs. work demos: what is the difference?

A work sample is the finished output: a file, a link, a document.

A work demo is the finished output plus a recorded screenshare in which the candidate shows how they made it and explains their decisions, ideally while answering follow-up questions.

Why work demos are better

• They verify authorship. It is easy to submit AI-generated or borrowed work. It is much harder to explain every decision in it on the spot.
• They show reasoning. Two candidates can produce similar outputs for very different reasons. The explanation shows which one understands the trade-offs.
• They combine two strong methods. A work demo pairs a work sample with consistent, job-related questions, which is the core of a structured interview. Selection research generally finds that combining multiple methods predicts performance better than any single method.
• They save interview time. Reviewers can watch the walkthrough when convenient instead of scheduling a separate presentation round.

This is the model Pudding uses. Candidates submit a deliverable and then explain it in a screenshare with an AI agent that watches and asks questions about their decisions, while humans on the hiring team judge the result. Hiring teams write their own brief, can choose between deliverable-plus-explanation or explanation-only, and can anonymize submissions before review.

How to pay candidates for work demos

Why pay

Payment signals that you value the candidate's time and increases the number of strong candidates who complete the task. Coderbyte notes that top candidates often have multiple offers and little reason to spend time on long assignments, and that paying for longer projects can improve completion rates among them. Basecamp has paid design candidates for projects of 10-20 hours, treating them like a small freelance job whether or not the candidate is hired.

Payment is not a complete fix on its own. A task that is too long, vaguely briefed, or followed by silence will lose candidates whether or not it pays. Short tasks, clear instructions, and fast feedback still matter.

A simple way to set the amount: estimate the hours the task should take and multiply by the role's approximate hourly pay. Use a flat amount so every candidate is paid the same.

What makes manual payment difficult

Paying candidates by hand means:

• Collecting payment details from each candidate
• Collecting tax forms where required
• Sending payments across different countries and currencies
• Tracking who has been paid and answering status questions
• Recording payments for tax reporting

In the US, the reporting requirement is lighter than it used to be for small honorariums. Starting with the 2026 tax year, businesses only need to report payments to non-employees on Form 1099 once the total reaches $2,000 in a calendar year, up from $600. Check with your accountant for your specific situation.

Use a platform that automates payment

A work sample platform with built-in payments removes these steps. In Pudding, every completed work sample is paid automatically through Stripe, and candidates get a dashboard to track earnings, payout status, and history. Candidates also keep full ownership of what they produce, whether or not they are hired. Pudding integrates with Greenhouse, Ashby, Workable, Lever, and RecruiterFlow.

Frequently asked questions

What is work sample automation?

Using software to send work sample requests, manage deadlines and reminders, collect submissions, record candidate explanations, and pay candidates, instead of handling each step by email.

How long should a work sample take?

Most teams aim for a few hours. Longer tasks should be paid and reserved for finalists.

Should you pay candidates for work samples?

Yes, for anything beyond a short exercise. The main reason is practical: asking a strong candidate to work for hours for free reads as an insult, and the candidates with other offers are the ones who walk. Paying keeps them in your funnel, which is why Coderbyte reports better completion rates on paid projects and why Basecamp pays design candidates for projects of 10-20 hours whether or not it hires them. The second reason is that it is the right thing to do, and where the task produces work the company could use, unpaid labor can also run into wage rules.

Can a company use a candidate's work sample?

It should not. Design the brief so the result is useful to the candidate but not to your business: avoid your own product, roadmap, customers, and internal data, and state in writing that the candidate keeps ownership.

How do you handle AI use in work samples?

Decide which AI tools are allowed and say so in the brief. Then ask candidates to explain their work in a recorded walkthrough. A candidate who understands the work can explain it, whatever tools they used.

What roles benefit most from work demos?

Roles without public portfolios, such as sales, product management, finance, and operations, and any role where AI can easily produce a convincing finished artifact.

Sources

1. Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology. Psychological Bulletin, 124(2). https://doi.org/10.1037/0033-2909.124.2.262
2. Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. Journal of Applied Psychology, 107(11). https://pubmed.ncbi.nlm.nih.gov/34968080/
3. SIOP, "Is Cognitive Ability the Best Predictor of Job Performance?" https://www.siop.org/tip-article/is-cognitive-ability-the-best-predictor-of-job-performance-new-research-says-its-time-to-think-again/
4. Anderson, N., Salgado, J. F., & Hülsheger, U. R. (2010). Applicant reactions in selection. International Journal of Selection and Assessment, 18. https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1468-2389.2010.00512.x
5. Monster, Job Search Deal-Breakers Report (2026). https://www.monster.com/career-advice/research/job-search-dealbreakers
6. Gartner, candidate AI use and fraud survey. https://www.gartner.com/en/newsroom/press-releases/2025-07-31-gartner-survey-shows-just-26-percent-of-job-applicants-trust-ai-will-fairly-evaluate-them
7. GitHub, Octoverse 2025. https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/
8. Jason Fried, "Hiring programmers with a take-home test," Basecamp. https://basecamp.com/articles/hiring-programmers-with-a-take-home-test
9. The Holloway Guide to Technical Recruiting and Hiring, "Take-homes." https://www.holloway.com/g/technical-recruiting-hiring/sections/take-homes
10. Coderbyte, "Compensating candidates for take-home projects." https://medium.com/coderbyte/compensating-candidates-for-take-home-projects-e10290c00a18
11. InHerSight, "Are Unpaid Take-Home Interview Assignments Ethical?" https://www.inhersight.com/blog/interview/are-unpaid-interview-assignments-ethical
12. UX Playbook, "How to Present NDA Design Work in Your UX Portfolio." https://uxplaybook.org/articles/ux-design-portfolio-nda-guide-2026
13. Ashby Recruiting Operations Benchmarks (2026), as summarized by Four-Leaf. https://four-leaf.ai/blog/unpaid-take-home-assignment
14. OnPay, "1099 Reporting Thresholds: 2026 Changes Explained." https://onpay.com/insights/1099-reporting-threshold-updates/