Airtable bases decay quietly: empty required fields, the same client entered twice, a date in the wrong year. An Orkeste agent reads a table on the schedule you set, lists what looks wrong and proposes fixes. An ops lead approves each fix before the record is updated.
The problem
Errors are found when a report looks wrong or a client receives the wrong email. By then, they have spread to everything built on the table.
How the workflow runs
Runs: Every day or every week.
Read the records: The agent reads the records of the table named in its briefing.
Check the rules: It checks required fields, formats, value ranges and likely duplicates against your rules.
Report: It posts the list of problems to Slack.
Propose fixes: Where the right value is clear, it prepares the update.
Approve: Each fix waits in Validations for the ops lead.
Update: Approved fixes are written to the records.
Every fix waits in Validations; nothing changes in the table without approval.
Who does what
Data checker: Finds and reports problems in the table.
Ops lead (human): Approves every fix.
What this workflow does not do
Updates records; it does not delete or merge duplicates. A person removes them.
Does not change fields, views or the table structure.