Which record types should I request in a generated schedule pack?
A checklist of record types and fields an agent team can ask for so that a generated schedule pack supports both training and grading of adjustment tasks.
Use this list to specify a pack so that it supports the actions your agent will take, and the checks you will run afterwards.
- **Contracts or accounts** with product code, start date, term and frequency. - **Schedule versions** with status, `supersedes_id` and the reason for each change. - **Instalment rows** with due date, amount, state and the payment that settled each one. - **Payment events** including early, late, part, duplicate and reversed payments. - **Instructions and reminders** linked to the schedule version that created them. - **Requests and cases** with channel, requested change, approver role and outcome. - **Policy tables** covering limits, thresholds, calendars and fee rules, plus the text an agent would read. - **Task definitions** with starting state, expected end state and the refusal or escalation outcomes.
Also ask which fields are deliberately left inconsistent to create repair tasks, and whether those are labelled so you can separate them from clean training examples. A pack that mixes clean and broken state without labels is hard to grade.
Naming record types up front keeps a pilot focused on the links that matter for your agent's actions.
- Audience:
- Agent developers
- Content type:
- Checklist
- Buyer question:
- What can be generated
Content release: r-a220eb6e099b243b
Related questions
- Can a release include policies, exceptions and refusals, not just records?
- What file layout and keys would a schedule dataset use?
- What would a simulated utility payment-plan environment hold?
- What can a generated schedule dataset or environment contain?
- Which entities should I ask for in a schedule dataset or environment?
- How could schedule records be laid out in JSONL and Parquet files?