What the work actually involves

Day to day, you are working through financial tasks the way you would for a client — then judging how well a model did the same thing. A typical batch might ask you to review a set of financial statements for GAAP or IFRS conformity and mark where a model's assessment was wrong; to build a working Excel model (revenue build, deferred tax schedule, lease liability amortisation) that serves as a reference answer; to document an audit procedure and the internal control it tests; or to write a tax compliance memo on provided facts. A large part of the output is written: findings, reasoning, and a clear explanation of why a model's answer fails, not just that it does.

You will also be asked to contribute scenarios from your own experience — the ambiguous revenue recognition call, the control deficiency that only surfaced in walkthrough, the disclosure that technically complied but misled. These become training and test material, so they need enough factual scaffolding that another accountant could work them independently.

What the screening looks for

micro1's screen is AI-led and conversational, with follow-up questions that push on whatever you claim. It is trying to confirm three things: that your reporting, audit, or tax experience is hands-on rather than supervisory-at-a-distance; that you can cite standards and treatments precisely under pressure rather than gesturing at them; and that you can explain a technical judgment in plain, structured prose. Expect to be asked about specific standards, materiality reasoning, and how you would grade a plausible-but-wrong answer.

  • Depth is tested by follow-up, so name the standard, the entity type, and the year when you can.
  • Writing quality matters as much as technical accuracy — vague findings are the most common rejection reason.
  • CPA, ACCA, CA or equivalent is valued but stated as preferred, not required; five-plus years of substantive experience carries similar weight.

Logistics

Fully remote, contractor, asynchronous. Work arrives in batches with turnaround windows rather than fixed shifts, and volume fluctuates with the customer project's needs — many contributors treat it as 10–20 hours a week alongside existing practice. Excel is essential; you will be building and checking models, not just reading them. Pay in the $60–75/hour range has been observed for this role and is not guaranteed for any individual engagement.