Overview
SPX uses AI to help engineers generate, improve, summarize, and reason about digital twin work, but AI output should always be treated as something that needs review.
Responsible AI for SPX means keeping humans in control of the final decision, especially when AI affects production behavior, security, or customer-facing systems.
Where AI Helps
AI can help draft models, suggest structure, summarize project context, highlight missing information, and speed up repetitive work that would otherwise take longer to perform by hand.
The goal is to reduce manual effort while keeping the engineering team in charge of the outcome.
Human Review
Engineers should review AI-generated output before relying on it in a validation flow, a production workflow, or any customer-facing system.
If the AI output conflicts with observed behavior, domain knowledge, or system requirements, the human review should win.
Data Handling
AI features should only use the input needed to complete the task at hand, and they should avoid exposing unrelated project details when that is not required.
If a feature relies on a third-party model provider, the feature should be designed with the minimum necessary context and with the protections available for that provider.
What AI Should Not Do
AI should not be treated as an autonomous operator for production systems, and it should not bypass human approval when that approval is required.
- Do not trust AI output without review
- Do not use AI to replace domain validation
- Do not treat AI suggestions as guaranteed facts
- Do not let AI bypass access controls or policy checks
Provider Changes and Limits
AI capabilities, model choices, and provider settings may change as the product evolves.
When that happens, we will aim to keep the behavior understandable and to document any material change that affects how AI features should be used.
Contact
If you have a question about an AI-assisted workflow or think an AI feature should behave differently, contact us at the address below.