The catch that pays for this is dull: the same amount, on the same day, twice on one report, one paid by card and one marked cash. Codex is good with ids, dates and amounts, which is what duplicate detection reduces to once somebody actually looks. The interesting failures here are arithmetic rather than judgement.
Run a staged conversation — no account needed. The agent handles it for real while a simulated world answers its tool calls; nothing touches real accounts, and nothing is actually sent.
Berlin Sales Trip, 4 to 7 May
Eleven line items totalling 1,842.60 EUR. Flights Warsaw to Berlin at 312.40 EUR, hotel for four nights at 688.00 EUR, seven meals between 14.20 and 61.90 EUR, and airport taxis at 74.00 EUR. Every item has a receipt attached and each dinner sits under the 75.00 EUR limit.
Using this template drops you into a guided setup. It asks exactly this, nothing else:
Connect Expensify
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect Slack
One sign-in. The agent acts through your account, scoped to what this template uses.
Expense policy
The spending rules a report has to satisfy - per item and per category limits, what always needs a receipt attached, categories you never reimburse, how far back a submission can be dated, and what makes a report worth a proper read rather than a rubber stamp.
Runs on Codex
Preselected for this page — connect your Codex account during setup, or switch to NoClick's built-in models with one click.
Watch it handle a test run
A staged conversation against a simulated world — then it’s live.
Duplicates hide in reports that are otherwise ordinary, and nobody clicking through forty of them on a Friday finds one. A structured pass over the line items does.
A conference that ended in March and a submission dated July is a policy question, and it surfaces only if the dates on each line are read rather than skimmed.
By comparing amounts, dates and merchants across the line items of that one report. It cannot see across reports or across months, so a duplicate split between two submissions is outside what it catches. Knowing that limit matters before you treat a clean verdict as an audit.
Free to start. Guided setup, a test run against staged conversations, and it's live.