Press Ctrl+J, or click Ask at the top of the sidebar, and choose Hazard Log Assistant under the message box.
2
Describe the hazard and name the product
Write it as you’d say it to a colleague. For example:
Add a hazard to the Clinic Triage Assistant product: an urgent red-flag symptom is not detected, so the patient is offered a routine appointment instead of same-day clinical review.
Click Send.
3
Check the proposal
The assistant proposes the hazard with a suggested harm, and asks what to set before it writes anything.
The assistant's proposal for the Clinic Triage Assistant. Nothing has been written yet: it lists what it hasn't set and asks three questions first.
4
Confirm, and add the detail you want
Answer its questions in one message, as in the screenshot below: the ID, any scores, and the cause and control to add. The assistant creates them in the same log you’d edit by hand, so they appear straight away.
What the assistant created: a hazard (HAZ-029), a cause linked to it (CS-029) and a control linked to the cause (CTRL-047).
5
Open the hazard log and check it
In Clinical Safety → Hazard Log, open the new hazard under its product and check its wording, scores and links before anyone relies on them.
The assistant turns words like Medium and Major into numbers on the log’s scale, so check the rating matches what you meant.
The control it creates has no evidence attached yet.
The assistant only drafts: your clinical safety officer decides whether a risk is acceptable.
Ask for a change the same way, naming the hazard and product:
In the Clinic Triage Assistant product, set HAZ-029’s residual likelihood to Low and severity to Considerable, and wait for my confirmation.
The assistant shows each field’s current and proposed value, and changes nothing else. Once you confirm, it saves the change and recalculates the residual risk score.To reverse a change, ask it to undo it: it proposes the old values and restores them once you confirm. Undoing a residual rating clears the residual likelihood, severity and risk.
Delete a hazard, cause or control. Asked to delete a control, it offers to unlink or archive it instead.
Approve a new version of the log or publish it to your Trust Centre.
Work on clinical documents. Today it works only on the hazard log.
To import an existing Excel or CSV hazard log or update many hazards at once, connect Claude, ChatGPT or another assistant through the Assuric MCP Server with Editor access.
Asked to delete a control, the assistant offers to unlink or archive it instead, and warns that the cause would be left with no control.
The same mode answers questions about the log; say “without changing anything” to keep it read-only. In the example below, likelihood is multiplied by severity, so scores run from 1 to 25.
A read-only question. The assistant lists the hazards that match, with their ratings and statuses, and changes nothing.
