What does SafeKivi AI actually do?
It works in three layers. Inside the modules it drafts and suggests: investigation summaries, SDS fields, import mappings and risk analyses. For analytics it turns a plain-words request into a report and summarises the result. And the SafeKivi connector lets Claude, ChatGPT or Microsoft Copilot answer questions from your live SafeKivi records.
Does AI change our records on its own?
No. AI produces drafts and suggestions. A person reviews and approves them before anything is saved, and the connector is read-only.
Who decides what data AI can see?
Your administrators. Each site has one of four AI data-access levels: Off, Metadata only, Aggregates only or Row level. Within that level, AI sees only what the signed-in person is allowed to see, and sensitive fields stay out unless you allow them.
Can we prove what was sent to AI?
Yes. Every AI request is logged, including refused ones, and SafeKivi produces an AI data-processing statement from your actual settings, with your recent usage, for your compliance team to keep on file.
What is MCP?
The Model Context Protocol is an open standard that lets AI assistants such as Claude, ChatGPT and Microsoft Copilot connect to other software and ask it questions in real time, within each person’s permissions.
Do we need IT to set up the connector?
An admin switches it on and chooses who may use it. After that, each person connects their own AI tool in a few minutes by signing in to SafeKivi.
Does the connector replace the SafeKivi screens?
No. Dashboards, forms and workflows stay as they are. The connector is another way to read the same data.
Can SafeKivi AI be switched off?
Yes. Set a site to Off and no AI processing happens for it at all. You can also turn individual AI features on or off.