User Guide
You use the copilot in three ways: ask · propose · automate. The sidebar lists New Chat · Proposals · Automation from the top, and asking starts from New Chat and the home screen. What lives where, and where the old menus went, is laid out on a single page in Copilot Screen Map.
PlantPulse AI consists of seven components. This section covers what each component does and how to use it. If this is your first time, it is faster to start with the concepts — Product Vision · Platform Architecture.
Where people work
| What you want to do | Document |
|---|---|
| Open the copilot — web · home screen app · Windows app | How to Open the Copilot |
| See where the sidebar (New Chat · Proposals · Automation), right panel, notifications, and settings are | Copilot Screen Map |
| Take the to-dos, anomalies to check, and incidents found by AI from approval → action → result → recurrence check | Proposals |
| Hand off your daily checks in a single sentence — results come back as proposal cards and reports | Automation |
| Ask and get answers in conversation — text · images · speech | AI Chat Web |
| Get answers grounded in a specific spot on a drawing | Asking with Drawings |
| Get answers grounded in a specific scene from maintenance or training video | Asking with Video · Audio |
| Understand how an anomaly is detected, becomes an incident, and gets resolved | Anomaly Detection · Incident Response |
| Mark whether the problems AI found were correct, and review accuracy | AI Findings and Accuracy |
| Ask about job readiness (Go/No-Go) and customer PO headroom on an installation with a connected digital thread | Digital Thread — Job Readiness |
| Define who does what and when | Work Process |
The old screens «Today · Anomaly Diagnosis · Resolution · Reports · Dashboard» are gone. The same answers come from asking in Ask, and items to handle gather under Proposals → Copilot Screen Map — Old Addresses
Where answers are built
What the AI «knows» comes from three sources — documents (RAG), the plant's structure (ontology), and time series values (TimeSeries-Insight). If an answer looks wrong, start by checking which of the three is empty.
| What the answer is grounded in | Document |
|---|---|
| Finds and answers from technical documents | RAG |
| Knows the plant's structure and relationships | Ontology |
| Analyzes time series | TimeSeries-Insight |
| Pulls in real-time data | MCP |
| Which model answers | AURA LLM |
Going deeper
How the algorithms work is covered in the Appendix. For integration, see the API Integration Guide.