A useful answer shows its evidence and what is missing.
Before preparing an intervention, your teams need to find the elements used, understand analysis limits and validate the next step. The AI offer connects your assistants to industrial context accessible within the agreed scope.
A concrete example
P-204: prepare an investigation without confirming the cause.
A flow discrepancy guides analysis. On its own, it does not demonstrate why a mechanical seal fails repeatedly.
Business question
Why has pump P-204 been repaired three times since January?
- History consulted
- CMMS history records three mechanical seal replacements. It documents recurrence, not its cause.
- Measurement and reference
- Average measured flow is 85 m³/h. The manufacturer manual specifies a minimum of 120 m³/h. The discrepancy calls for examining hydraulic conditions.
- Hypothesis
- Cavitation or recirculation to investigate. The answer presents these as hypotheses, without turning them into a confirmed diagnosis.
- Missing information
- Suction pressure is not instrumented. An on-site measurement is needed to continue the investigation.
- Proposal to validate
- Prepare a hydraulic investigation with plant teams. They review the evidence and define checks before any operational next step.
Editorial example taken from the P-204 case already presented on the site. It illustrates how to read an answer and is neither a client result nor a live AI interface.
How it works
Retrieve context, then verify the proposal.
- 01
Consult planned sources
The assistant accesses information authorised for the defined role and scope: history, measurements and documentation relevant to the question.
- 02
Separate evidence levels
Measurements, documents, calculations and hypotheses are distinguished. Insufficient evidence and missing information explicitly limit the conclusion.
- 03
Keep human decision-making
The assistant prepares a proposal. Transmission to the CMMS requires human validation and the connector planned for that action.
Scope and status
Access and commissioning defined with you.
- Controlled access
- Roles, equipment, sources and connectors are scoped with your organisation. The P-204 example only uses history, measurements and the manual available within that scope.
- Logging
- Queries are logged so exchanges can be retrieved. Log content, access permissions and retention periods are specified with IT.
- What you buy
- AI licences for the required users, integration, commissioning and onboarding within the agreed scope. The offer can extend REX. Sources, uses, support and acceptance criteria are set out in the proposal.
- Hosting
- Defined with you according to plant constraints. The Sovereign package provides the server and model in your infrastructure, with no calls to an external AI provider.
Which question are your teams trying to resolve?
Bring a business case and the available sources. We define uses, users, access and commissioning conditions with you.
Back to the AI offer