Can we adapt a generated pack for customers with different follow-up policies?
Customer-specific policy changes can alter correct actions, so resellers should scope adaptation rights and label regeneration together rather than treating customisation as text editing.
**Can a reseller request customer-specific policy variants?**
A buyer can ask MirrorFirm to scope variants around different triggering conditions, escalation roles or required evidence. Whether the reseller may create and distribute its own variants is a licensing question to settle explicitly.
**Why is replacing the policy document insufficient?**
The policy may control task labels, permitted actions and terminal states. Changing a waiting condition can turn an expected reminder into an instruction to wait. Leaving the old answer attached creates contradictory training material even when every record remains syntactically valid.
**What should a variant request include?**
Specify the policy differences, the affected tasks and the expected treatment of existing evaluation results. Ask whether derived labels will be regenerated, reviewed or marked unsupported. Preserve the relationship between the base release and the variant so downstream customers know which policy produced their material.
**What is a sensible boundary?**
Begin with named policy parameters that the pack is designed to vary. A request to substitute an entirely different process should trigger a fresh scope discussion. This keeps customer adaptation concrete and avoids presenting an unreviewed derivative as equivalent to the original generated product.
Preserve task validity by treating policy adaptation as a change to both the inputs and the evaluation expectations.
- Audience:
- Licensed resellers
- Content type:
- FAQ
- Buyer question:
- Scope and getting started
Content release: r-a220eb6e099b243b
Related questions
- What integration contract should be agreed before connecting a pilot to an agent harness?
- Which licence questions matter when generated material enters an agent development pipeline?
- What should developers include in a request for a generated sample?
- What makes a useful first pilot for an overdue-handling agent?
- How can a sample review uncover a missing requirement before a pilot starts?
- What integration contract should we agree before a pilot is generated?