Incident reporting for a care organisation

An assistant that helps staff report incidents completely and consistently, without taking the decision away from them.

Field
AI assistants, process automation
Client
A care organisation (not named)
Environment
The organisation’s own Microsoft 365, no external server
Status
Prototype presented in October 2026, in development

The task

Reports that are complete the first time

Staff know what happened. Classifying an incident, and knowing which details belong in the report, is a separate task where things get left out. The assistant takes over that part as far as it can.

How it works

Four steps, one decision by a person

INCIDENT INOWN WORDS FOLLOW-UPQUESTIONS PROPOSAL WITHREASONING PERSONCONFIRMS INCIDENT IN OWN WORDS FOLLOW-UP QUESTIONS PROPOSAL WITH REASONING PERSON CONFIRMS
  • Rule-basedEvery proposal is traceable.
  • Inside Microsoft 365No external server.
  • Person decidesOnly a confirmed report counts.
Details of each step
  1. Describe. A staff member writes down what happened, in their own words.
  2. Follow-up questions. The assistant asks targeted questions about whatever is still missing.
  3. Proposal with reasoning. It proposes a classification and shows why it arrived there.
  4. A person confirms. The proposal only counts once a person has checked and confirmed it, or changed it.

Rule-based

The classification follows fixed rules, so the reasoning shows how a proposal came about and anyone confirming it can check it.

Inside Microsoft 365

Reports don’t leave the environment the organisation already controls.

The assistant prepares. A person decides.

Status

Where the project stands

Prototype presented in October 2026. In development, not yet in productive use, so there are no results to report yet.

More on AI assistants Discuss a similar process