What the work actually looks like

You are given a prompt an accounting professional might realistically send a model — "reconcile this intercompany balance," "prepare the deferred tax rollforward," "assess whether this arrangement contains a lease under ASC 842" — along with one or more model responses. Your job is to decide whether the answer is right, and if it is wrong, to say precisely where the reasoning broke: a misapplied recognition criterion, a debit and credit that balance but land in the wrong accounts, a correct standard cited for the wrong fact pattern. Much of the value sits in that last category, because models are fluent at quoting ASC and IFRS paragraph numbers while misreading the transaction.

Tasks vary by project. Some weeks you are writing original problems with a fully worked gold-standard solution and a rubric another reviewer could apply. Others you are ranking two responses head to head, or rewriting a nearly-correct answer into the version a senior manager would sign off. A minority of projects touch adjacent ground: Excel model auditing, close-process narratives, audit workpaper review, indirect tax.

What the screen is looking for

micro1's intake is an AI interviewer that asks about your background and then follows up on specifics — which standard, which entity type, which year, what you did versus what your team did. It is testing whether the accounting knowledge is load-bearing or memorised. Beyond technical depth, it looks for evaluation judgment: can you separate an answer that is wrong from one that is merely different from how you would have done it, and can you write a critique someone else could act on. Vague, confident answers score worse than narrow ones with a clear scope.

  • Public accounting or senior in-house experience, with a licence (CPA, CA, ACCA, CPA Canada or equivalent) strongly preferred
  • Working command of US GAAP, IFRS, or both — stated honestly, not claimed for both when you live in one
  • Willingness to show your work in writing; most of the deliverable is prose reasoning

Logistics

Fully remote and predominantly asynchronous. Most contributors work 10–25 hours per week against a weekly task queue with rolling deadlines rather than fixed shifts, though some client projects require a few overlapping hours with US business time for calibration calls. Engagements are contract-based and scoped per project; volume fluctuates, so treat this as a supplement rather than a replacement for a full book of work. Pay is stated as observed across recent placements and varies with specialisation, review tier, and client — it is not a guaranteed rate.