Skip to content
MirrorFirm

What should we ask about samples, licensing and training use before committing?

Answers the common buyer questions on what a sample can contain, how licence terms are set, whether data can be used for training, and what is not promised.

**What can we ask for in a sample?**

A buyer can ask for a small linked set covering your chosen domain: the manifest, a handful of schedules with versions, events, policies and tasks with expected end states. Ask for at least some cases that end in refusal or escalation, and some that start mid-case with handover notes.

**Can we choose the policies?**

Yes, that is a scoping point. You can supply the rules you want tested, or ask for generated policies and review them for fit.

**How is licensing handled?**

Licence terms are agreed per engagement. Questions to raise include permitted uses (evaluation, training or both), internal versus external sharing, publication of benchmark results, retention, and what happens to derived artefacts such as fine-tuned models. This entry sets no terms and quotes no prices.

**Can we train on it?**

That depends on the licence you agree. Keep evaluation tasks separate from training tasks, and decide up front which variants are held out so results stay meaningful.

**What is not promised?**

The output is generated. It does not come with any guarantee of anonymity, privacy compliance or regulatory approval, and it is not real business records. Ask how the output was grounded, and inspect a sample for disclosure risk yourself.

**What is a sensible first step?**

A one-page brief, then a small sample reviewed by someone who does the work.

Asking these questions early avoids surprises about permitted use and sets realistic expectations for what generated data can support.

Audience:
Evaluation teams
Content type:
FAQ
Buyer question:
Scope and getting started

Content release: r-a220eb6e099b243b

More on Repayment scheduling