Virtual Assistant Trial Projects: Test the Real Job Without Live Risk

A portfolio and an interview answer different questions

A candidate can describe excellent organization without showing how they handle an ambiguous request, preserve a source, or escalate a risky exception. A portfolio shows past output, but it may reflect a different tool, reviewer, or level of support. A short trial project can add job-relevant evidence when it is designed as a sample rather than a disguised shift of unpaid work. Start with the role charter. Identify two or three behaviors that matter in the first month: perhaps classifying an inbox against written rules, turning meeting notes into an action register, or checking product changes against an approved request. Do not create a generic puzzle merely because it is easy to administer. Typing speed has little value if the job succeeds through careful exception handling. The sample should be short enough that candidates with current work and caring responsibilities can participate. Pay for substantial assignments and state the amount and payment route before work begins. Never publish, send, sell, or otherwise use a candidate's output as production unless a separate, explicit arrangement permits that use. The hiring purpose and the business-production purpose should not blur.

Remove live risk while preserving real judgment

Rebuild the task with synthetic or properly redacted records. A support trial might contain eight fictional tickets: one routine answer covered by the knowledge base, one duplicate, one customer asking for an unauthorized refund, one suspected account-security issue, and several ordinary cases. Ask the candidate to classify each item, draft only the responses allowed by the rules, and create an escalation note for the rest. Preserve the structure that makes the work difficult. Include dates, conflicting fields, an outdated note, and a missing approval where those are normal conditions. A perfectly clean dataset measures compliance with an obvious path rather than operational judgment. At the same time, do not plant tricks with no job relevance. Candidates should be able to find the controlling instructions. Keep the exercise outside live email, CRM, accounting, applicant tracking, and ecommerce systems. Use a sandbox or static packet, remove secrets and personal data, and disable external sending. If a tool simulation is necessary, give every candidate the same access and setup time. Document whether outside research or AI tools are permitted and which information may not be entered into them.

Write the brief like a real handoff

A strong brief states the business situation, desired output, available sources, authority boundary, time box, submission format, and who receives questions. It distinguishes facts from assumptions. For example: “Prepare a proposed Tuesday schedule from these requests. Do not move the marked client call, accept fees, or contact attendees. List conflicts and questions in a separate note.” Give candidates a reasonable question path. The quality of a clarification can be evidence, especially when the role regularly receives incomplete requests. Record the answer and share it with every active candidate if it materially changes the task. Otherwise the assessment begins measuring who happened to ask first. Include a stop rule. A candidate who finds exposed personal data, an instruction conflict, or an action outside the stated authority should know to pause and report it. In many assistant roles, recognizing when not to proceed matters as much as producing a polished artifact.

Build the scorecard before seeing names

Choose four to six dimensions tied to performance. For an operations sample, they might be accuracy, source fidelity, prioritization, boundary recognition, completeness, and communication. Define what weak, acceptable, and strong evidence looks like. “Good judgment” is too vague; “routes the refund exception to the named owner and cites the conflicting order evidence” is observable. Weight critical errors separately. A beautiful schedule that reveals confidential notes or commits the founder without authority should not pass because its formatting is excellent. Likewise, do not over-penalize cosmetic choices that the organization can teach quickly. Score the artifact before discussing personal style, and have reviewers cite evidence from the submission. Use the same core task, instructions, time allowance, and rubric for candidates being compared. Provide reasonable adjustments through a clear route. The US Equal Employment Opportunity Commission advises employers to ensure selection procedures are job related and do not unlawfully discriminate; its [employment tests guidance](https://www.eeoc.gov/laws/guidance/employment-tests-and-selection-procedures) is a useful starting point. Local employment rules still need qualified review.

Close the trial respectfully and learn from it

Acknowledge receipt, explain the decision timetable, and provide the promised payment promptly. Store submissions only as long as the hiring process and applicable policy require. Limit access to the hiring team, and do not add candidates to marketing lists because they completed an assessment. When possible, offer concise feedback anchored to the rubric: the escalation choices were strong, but two source discrepancies were silently normalized. Avoid presenting subjective preferences as universal truths. If several capable candidates misunderstand the same field, improve the brief rather than concluding that the market lacks attention to detail. Review whether trial performance predicts onboarding outcomes. After a hire's first month, compare the sample dimensions with supervised work. Remove criteria that add burden but no useful signal. A trial is a selection tool, not a rite of passage. Before reusing the exercise, audit whether its inputs or expected answer have gone stale. A calendar sample built around an old meeting policy may reward the wrong choice after scheduling rules change. A CRM sample can become misleading when required fields, ownership rules, or consent handling change. Give the task and rubric an owner, version, and review date. Keep a clean master copy, then record which version each candidate received so reviewers do not compare submissions against different standards. Also inspect the exercise from the candidate.s side. Confirm that every linked file opens without requesting personal accounts, every instruction is accessible in the promised format, and the submission route does not expose one candidate.s work to another. Test the time box with someone who understands the role but has not seen the sample. If they spend most of the allotted time deciphering the setup, the exercise is measuring familiarity with the test designer rather than readiness for the job. For a broader view of fair assessment, consult the US Office of Personnel Management's [assessment and selection resources](https://www.opm.gov/policy-data-oversight/assessment-and-selection/) alongside the EEOC guidance. If you want help defining a safely scoped assistant role before testing candidates, [contact us](/contact) or review our [operations assistant services](/services/operations-assistant-staffing).

Provider questions to copy

"Can you show how this role is screened, trained, checked each week, and replaced if fit is poor?"

"Can we start with a small task list before we expand the role?"

FAQ

Sources and notes

These sources are included as planning references. They do not replace legal, tax, security, or HR advice.

• EEOC Employment Tests and Selection Procedures: Official US guidance on job-related and non-discriminatory selection procedures.
• OPM Assessment and Selection: Official resources on structured employment assessment.