Two candidates describe themselves as organised and detail-oriented. One gives an impressive interview; the other asks careful questions about your instructions. A hiring decision needs evidence of what each person can do in the actual role.
For a Philippines-based virtual assistant, assess the individual’s relevant skills, judgement and communication. Nationality, accent and interview confidence are not substitutes for that evidence. A short, structured interview and paid fictional work sample can help you compare the capabilities the role needs.
The scorecard below is an original worksheet to adapt before meeting candidates. It supports a human hiring decision; it is not a validated selection instrument or a promise of future performance.
Choose the criteria before the candidates
Start with the role’s accepted outputs and responsibilities. Our first VA role-design guide helps define those. For this example, assume a service administrator who maintains job records and prepares routine status updates.
Choose criteria that the assessment can actually reveal: source checking, record accuracy, clear communication and recognising when a decision is needed. Separate essential starting skills from systems knowledge you will teach. Do not reject someone for unfamiliarity with your particular software if the role includes training in it.
Agree how interview evidence and the work sample will inform each criterion. The NSW Government’s public-service recruitment guide describes structured interviews using common questions and rating standards, and work samples linked to role requirements. Those assessment principles are useful here; its public-service rules are not presented as rules for a private VA engagement. Read the official guide.
Ask comparable questions and useful follow-ups
Give each candidate the same core questions. Ask for specific examples without requesting confidential material from previous clients:
- Describe a task where two sources disagreed. How did you establish what to use?
- Tell us about an error you found in your own work. What changed afterwards?
- Describe a request you could not resolve within your authority. What did you communicate?
- Walk through how you learnt an unfamiliar process and checked your understanding.
Follow-up questions can clarify the individual’s contribution: what they checked, what they personally did and what evidence showed the work was complete. Record those details rather than impressions such as “good attitude”.
Explain the role’s actual contact hours and communication channels, then discuss whether the arrangement works. Invite requests for assessment adjustments. Assess understandable, appropriate communication for the job rather than resemblance to the interviewer’s speaking style.
Agree a short paid sample with fictional records
Before the exercise, confirm its purpose, scope, expected duration, payment amount, payment date and payment method. Explain permitted tools, what assistance is available, how results will be used and whether the exercise will be recorded. Set terms appropriate to the proposed engagement; completing it does not imply an offer.
Provide the same instructions and resources, with agreed adjustments where needed. State the time zone for deadlines. Pay for the agreed participation regardless of the hiring outcome. Keep the exercise proportionate and separate from productive customer work.
A suggested sample is a 45-minute fictional job-update exercise. Supply six job records, three incoming notes and an approved status guide. Include a missing scheduling decision and one conflicting reference. Ask for an updated tracker, a short customer-update draft and a note listing unresolved questions.
Build all records specifically for training, without real customer data or live system access. Pilot the exercise with someone familiar with the role to check clarity and timing. If AI use is permitted, specify the approved tool and ask candidates to explain what they checked themselves.
Score the evidence against clear anchors
For each criterion, record the observed evidence and a score. Use NE when the assessment did not provide enough evidence; that requires clarification, not an automatic zero. These example anchors fit the fictional administrator role:
| Criterion | 0: not demonstrated | 1: partly demonstrated | 2: meets the stated standard |
|---|---|---|---|
| Relevant reasoning | Cannot explain their own actions | Describes actions but leaves checks unclear | Explains actions, checks and relevant results |
| Source checking | Treats an unsupported note as confirmed | Notices conflict but does not identify the missing decision | Traces the conflict and requests the necessary clarification |
| Work quality | Material record errors remain | Mostly accurate; specified corrections needed | Required fields accurate, with unknowns clearly marked |
| Communication | Misstates status or makes unsupported promises | Meaning understandable but next action unclear | Accurate message with a clear permitted next action |
| Decision limits | Acts beyond the supplied authority | Recognises a limit but leaves no useful escalation | States the issue, decision needed and appropriate owner |
| Response to feedback | Repeats the identified error without addressing it | Corrects the example with further prompting | Corrects it and explains the rule for future work |
For the final row, give everyone a comparable clarification or correction and a brief opportunity to respond. Do not reward candidates for guessing information deliberately omitted from the sample.
Resolve differences before making the decision
Where two reviewers are available, have them score separately before discussing evidence. If their scores differ, identify the particular field, statement or interview response behind the difference. Resolve ambiguous instructions before attributing the problem to a candidate.
A total can summarise the worksheet, but it should not override an essential criterion or conceal missing evidence. Agree those essentials beforehand. Do not change the scoring rules midway because one applicant seems especially likeable.
If the interview and sample suggest different capabilities, ask a focused follow-up. Distinguish a correctable training need from a starting requirement the evidence has not demonstrated. The responsible human reviewer records the decision and its job-related reasons.
Keep AI and candidate information within clear limits
AI can help draft generic questions or fictional instructions for human review. In this workflow it does not score, rank or select candidates. Do not upload candidate records to a public chatbot as a shortcut.
The OAIC’s recruitment example warns about inaccurate or biased AI outputs and calls for appropriate controls and human oversight. It also recommends keeping personal information out of public generative AI tools as best practice. Read the OAIC guidance.
Restrict access to assessment notes, explain their intended use and manage them under the business’s retention arrangements. For the selected candidate, turn identified training needs into an onboarding plan. Talk to Nexus about defining the role and assessing the practical work it requires.
