Conversation examples

Five real flows already running in production. Notice how Claude responds with concrete numbers and always leaves a decision in plain sight.

A · Daily reconciliation close

You — Load today's payments file, run the transaction reconciliation, and when it finishes, export the result to the close folder.

Claude — Uploading pagos-2026-05-21.csv to the collections source… file loaded. Launching the transaction reconciliation… finished with 4,872 matched and 18 pending. Exporting the result to the "cierre-diario-s3" connection… done, file available.

B · Ad-hoc investigation

You — How many pending transactions do I have from the last week in the collections source?

Claude — Querying the source… you have 247 transactions in PENDING status dated between 2026-05-14 and 2026-05-21. Do you want me to show the detail, export them, or group them by some criteria?

C · Set up a new reconciliation

You — I want to reconcile the "pagos-stripe" source against "ingresos-banco", matching on transaction_id and allowing 2 days of tolerance on the date.

Claude — Reviewing both sources… I found the columns. Creating the "Stripe vs Bank" reconciliation with the rule transaction_id = transaction_id and a date tolerance of ±2 days. I left it in draft — should I confirm and run it, or do you want to review it first?

D · Schedule tasks

You — I want this reconciliation to run every day at 7 a.m. Bogotá time, and to alert me if more than 50 transactions remain unmatched.

Claude — Scheduling the reconciliation for 0 7 * * * in America/Bogota and configuring an alarm with a threshold of 50 unmatched transactions. Done.

E · Audit and traceability

You — Which export runs ran yesterday?

Claude — Yesterday 12 export runs ran. 11 succeeded, 1 failed (export ID 890, connection error to the destination). Do you want to see the error detail or retry?