AI notice

Where AI is used

• Drafting use case descriptions, problem statements and expected benefits in AI Use Case Studio™.
• Suggesting scoring rationale for feasibility, data readiness and value — never the final score you commit to.
• Summarising a portfolio or roadmap into narrative sections of a report.
• Helping you phrase an opportunity captured from a workshop or conversation.

Where AI is never used

• Approving a submission into the public repository. Publication requires a human reviewer with an admin role.
• Granting entitlements, roles or workspace access.
• Deciding what your organisation should fund. The product records your decision; it does not make it.

What is sent to an AI provider

Only the fields relevant to the feature you invoked — for example the use case you are editing — plus our instructions to the model. Requests run server-side. Provider credentials never reach your browser.

• We do not send your workspace membership list, other workspaces' content, or your credentials.
• We do not use your private workspace content to train models, and we instruct providers not to.
• Inputs are validated and sanitised before they are sent, which also limits prompt-injection paths through pasted content.

Accuracy and human validation

• AI output can be confidently wrong, out of date, or generic. Treat every draft as a starting point.
• Check figures, benefit claims, regulatory statements and technical feasibility against your own sources before circulating them.
• The person who accepts a draft owns it. Audit events record who saved, scored, published or exported — not the model.
• Where AI drafted content, the product shows it as a draft you must confirm rather than saving silently on your behalf.

Fair-use metering

AI assistance is metered per user and per workspace over a rolling 24-hour window, with a short-burst limit to stop runaway automation. When a limit is reached, AI generation pauses until the window resets and everything you have already created stays available. Current limits are published on the Founding Access terms page and remain in force through 31 December 2026.

Reporting a bad AI output

If AI output is wrong, unsafe or inappropriate, report it from Help & Support. Include the surface you were on and what you expected. Reports are triaged with priority and tracked to resolution, and can be sent to support@usecasesx.ai.