Trust · Responsible AI
People decide when it's unclear.
AI is useful for understanding everyday speech. In a care setting, it also needs firm limits. These are ours.
Principles
Eight commitments.
- 01
People decide when it's unclear
Uncertain results go to a human reviewer. AskLight never turns uncertainty into a confident-looking task.
- 02
Narrow purpose
AI describes everyday requests. It does not diagnose, assess symptoms or decide on care.
- 03
The system routes, not the model
Room, unit, ownership and urgency come from your setup and staff actions — never from what was said.
- 04
No hidden listening
The resident starts the device. Audio is processed for the request and then deleted.
- 05
Works without AI
A manual help request needs no audio and no AI, so the most basic path never depends on a model.
- 06
Measured honestly
We measure misses, false alerts and review rate separately, test with realistic conditions, and don't present synthetic test results as real-world accuracy.
- 07
Traceable changes
Each result records the model and prompt version, so changes can be evaluated before and after.
- 08
Open about limits
English first. Real rooms vary. We say so — on this website and in the product.
Seen a result that looks wrong?
Supervisors can dismiss or correct requests in the review list, and every decision is recorded. If something looks systematically wrong, tell us through the contact form.
See AskLight on one of your units.
Tell us about your nursing home. We'll walk through how a pilot works and whether AskLight is a fit — no commitment.