What the work actually involves

You are handling the kind of question a US small-business owner brings to a chat window: why the bank feed won't reconcile, whether a payment to a contractor belongs in COGS or opex, how to unwind a prior year's misclassified owner draws, what the sales tax liability account is doing after a QBO migration. Some queries are shallow help-desk items. Many are data-heavy — you'll be reading P&Ls, balance sheets, transaction registers and AR aging, then writing an answer that is both technically correct and usable by someone who is not an accountant.

The AI component sits underneath. Your written reasoning becomes reference material and evaluation signal for models being trained on accounting and finance work, and in some workstreams you'll be comparing model outputs against what you'd have told the client — flagging where a plausible-sounding answer would produce a misstated ledger, a missed filing, or advice that crosses into tax or audit territory the model shouldn't be giving.

What the platform screens for

  • Advisory breadth, not a single seat. The listing is explicit: experience advising three or more organizations. Someone who has been controller at one company for a decade is a weaker fit than someone who has cleaned up books across a dozen SMBs.
  • QuickBooks Online at the hands-on level. Not "we used QBO" — chart of accounts design, bank rules, reconciliation discrepancies, class and location tracking, journal entry mechanics, the usual QBO failure modes.
  • US SMB context. Cash vs accrual conversions, S-corp owner compensation, 1099 handling, state sales tax, the practical realities of businesses under $20M revenue.
  • Willingness to say "this needs a CPA." Knowing the boundary between guidance and regulated advice is scored, not penalized.

Selection runs through the Vetsmith Challenge — a structured assessment — followed by client-side profile screening. Expect the challenge to include written responses to realistic finance scenarios rather than multiple choice.

Logistics

Fully remote and asynchronous in scheduling, though chat-support workstreams often carry coverage expectations within US business hours. Commitment is stated as up to 40 hours per week for 10 weeks, extension contingent on performance and project needs. Pay is not disclosed in this listing; Turing typically sets hourly rates per project and per seniority band, and rates observed on the platform vary widely by workstream — treat any figure you hear secondhand as unverified until it's in your offer.