Use case · Operations
Back-office automation.
An agent works through the routine paperwork on a schedule, while your team sleeps.
Checks your team does by hand get done every night. People only see the cases that need a decision.
How it works
Step by step.
Press play, or click through it yourself. Try the switch above the drawing.
Try it · Tonight
scheduleyesterday's workagentmatching codefixeda personmorning report
04:30
0104:30. The schedule wakes the agent. Nobody needs to be online.
01 / 06
Where it's used
One idea, many industries.
- 01FinanceMatch invoices to payments every night and refund double charges.
- 02ProcurementCheck supplier invoices against purchase orders and flag the gaps.
- 03OperationsClean up yesterday's orders and fix missing addresses.
- 04InsuranceCheck new claims for missing documents before an adjuster starts.
How Promptise makes it work
All three parts, working together.
In code
A peek at the code.
- 01The runtime keeps the agent running.
- 02The cron trigger wakes it at 04:30.
- 03Each tick becomes one run.
nightly.pypython
from promptise.runtime import AgentRuntime, ProcessConfig, TriggerConfig runtime = AgentRuntime()await runtime.add_process("reconcile", ProcessConfig( model="openai:gpt-5-mini", instructions="Reconcile yesterday's invoices and payments.", triggers=[ TriggerConfig(type="cron", cron_expression="30 4 * * *", cron_timezone="Europe/Berlin"), ],))await runtime.start_all()A real project you can run
Finance
Invoice reconciliation
An agent that reconciles invoices every night and follows up on failed payments, with refunds capped in code.
A reference application with local stand-ins for the outside systems.