Looking for work with a US company? Apply to the Rolemote talent roster — free →

Interviews reward fluency; the job rewards judgment. The questions below are scenario-first on purpose — past specifics and decisions under constraint predict, rehearsed strengths don't. Pair them with the graded work sample at the end and you have most of a real screening process.

1I'll describe five transactions — a software refund, an owner's personal charge, a Stripe payout, a refunded ad spend, a $1,900 laptop. Categorise each and tell me which one you'd flag instead of deciding.

What a strong answer shows: Mostly correct calls, and — the real test — the instinct to flag the ambiguous one (usually the owner charge or the capitalisable laptop) with a proposed answer rather than guessing silently.

2A receipt is missing for a $430 charge and the cardholder hasn't answered your first message. Walk me through your chasing process.

What a strong answer shows: Polite persistence with a system: a reminder cadence, an easy path for the cardholder (forward the email, snap a photo), an escalation point, and a record of the chase so nothing silently ages out.

3What's your personal system for staying accurate in hour six of repetitive work?

What a strong answer shows: Something real: batching with breaks, end-of-batch self-review, running totals that must reconcile, checking work at a different time than doing it. 'I'm just careful' predicts errors at volume.

4Where's your line between deciding yourself and asking?

What a strong answer shows: A threshold answer: routine and covered by the rules — decide; unusual, large, or tax-adjacent — ask, with a suggested answer attached so the question costs the reviewer ten seconds, not ten minutes.

5What does the first hour of your working day look like?

What a strong answer shows: A routine: yesterday's flags first, then the transaction queue, then chasing — some deliberate order rather than opening whatever shouts loudest.

6Tell me about the most repetitive job you've held. What did you like about it, honestly?

What a strong answer shows: You're screening for genuine temperament fit. Someone who found satisfaction in clean, finished, accurate work will stay; someone visibly bored describing it will be gone in four months.

7Do you use any automation or AI in data entry work? How do you check it?

What a strong answer shows: Comfort with bank feeds, rules and OCR — paired with spot-checking against source documents. Both 'I trust the tool' and total refusal are wrong answers.

Red flags, from reading a lot of these

Then test the work, not the talk

The single highest-signal step isn't a question at all. For bookkeeping assistants: A categorisation judgment set plus a receipt-chasing process. Graded work under a written rubric is how our own screening scores every candidate — here's a real sample shortlist showing what that evidence looks like.

Or let the screening run for you

Describe the role once and we run this whole process — sourcing, scenario answers, graded work samples, a second AI-tailored interview round — and hand you three scored finalists. The search runs free; the flat $1,995 is due only when you want introductions.

Hire a bookkeeping assistant — search runs free

Common questions

How many interviews should I run for a bookkeeping assistant?

Two, at most — one structured conversation built on questions like these, and one review of graded work. Adding a third round rarely adds signal; it mostly adds weeks, and in this market the good candidates are gone by then.

Should I send questions in advance?

Send the work sample in advance — polished written work under no time pressure is exactly what you're buying. Keep the scenario questions for the live conversation, where you can probe the first answer; the follow-up is where rehearsed answers fall apart.

Can't candidates just answer these with AI?

Written answers, sometimes — which is why the live follow-up matters, and why our own screening grades the thinking in an answer rather than its polish. Ask 'what happened next?' or 'why that order?' and a borrowed answer runs out of depth in one exchange. A candidate using AI well day-to-day is a positive signal; a candidate who can't defend their own answers is the flag.

What actually predicts a good hire?

Graded work beats conversation. Interviews reward confidence and fluency; a role-specific scenario plus a real work sample rewards the actual job. Use the interview to probe judgment and verify the work is theirs — not to guess at skill from charisma.

Related