CompanyBench

Audit manual-review check deposit release latency & stranded fundsDeposits & Funding

13 verifiers · 12 LLM-graded

EnvironmentsConfluenceJiraZeta3 SQL

TaxonomyDeposits & FundingDeposits & Funding

Task prompt

Hey, quick one but kind of important. We're drowning in 'where's my check money' tickets and I'm pretty sure a batch of our mobile check deposits just got stuck in processing and never actually paid out. Money never hit the customer, nobody released it, and it's just been sitting there. Can you dig in and tell me how bad this is? I only care about the older stuff, so anything submitted before July 2023. Newer ones might still be working through the pipe, so leave those out. I mainly need to know how many deposits are stuck like this, how much customer money is tied up in them, and how many separate accounts got hit. Then give me a feel for whether it's spread out or concentrated: what's the worst single account holding in stuck money, and how many accounts have more than one of these stuck deposits. There are two buckets I know about and I need them kept apart. Most of them are the ones that quietly died on the check-image capture and never got surfaced back to the customer. The rest are a smaller pile that got kicked to manual risk review and are just waiting on a human to release or return them. Tell me how many are in each bucket, and how much money is sitting in that risk-review pile. And how long has the worst one been dragging on? Take the single oldest stuck deposit and tell me how many days it's been sitting, from when the customer submitted it up to the review date, call it July 1, 2023. One thing before you trust that split: don't just eyeball the statuses. Check how our mobile check-deposit flow is actually supposed to work in the product docs, and see whether engineering already has a ticket open on this so we know the real root cause and I'm not re-reporting something they're already fixing. When you come back, walk me through the numbers and tell me what's actually causing these to strand and whether eng is already on it.