Work & OperationsDone with custom-built or unnamed agent Sep 16, 2026
Reconcile industry files and chase clarifications with agents
An agent workflow ingests a complex industry file, loads it into a database, spawns agents to check it against work in flight, and contacts people to resolve bad data.
- What They Did
- The author, building for a specific industry they don't name, has a $4/1M-token model ingest a file normally used in a complicated legacy program. It parses the data into a database, then spawns agents to check it against work in flight. Further agents contact the people involved to clarify bad data, and others keep reconciling against many items as updates arrive.
- How It Went
- The author says it replaces a $500+/hr service and catches mistakes in real time across languages and cultures. No measured results are given.
- Worth Knowing
- Expect patience: the author says it takes running tests and tweaking until it works, and their main ongoing job is maintaining the tests.
Try it yourself with custom-built or unnamed agent
I receive a [file type, e.g. industry data file] from [source] every [frequency]. Write a workflow that loads it into a local [SQLite or Postgres] database, then checks each record against [my in-flight work, such as a spreadsheet or table]. List every mismatch or bad record in a report, with a draft clarification email to the person responsible for each. Stop when the report and drafts are ready, and show me the emails before sending any.