Legal
Risk Disclosure
Last updated March 15, 2026. AI employees improve operational speed, but they do not remove business, legal, or workflow risk. This page outlines the major risk categories customers should understand.
Model and hallucination risk
- Nodebase explicitly bounds LLM capability through Retrieval-Augmented Generation (RAG) and strict semantic grounding.
- However, AI systems can still generate inaccurate or contextually poor outputs if the provided operating policies or reference materials are outdated.
- Customers must rigorously test the workflow with real operational edge cases before granting autonomous execution.
Operational risk
- Poor escalation design can allow messages or actions to cross the wrong threshold.
- Incomplete policy documentation can cause the employee to make inconsistent decisions.
- External systems such as messaging rails, calendars, and payment providers may fail or behave unpredictably.
Human responsibility
Customers remain responsible for setting policies, reviewing approval-sensitive actions, and ensuring that regulated or high-risk activities stay under human control where appropriate.
Nodebase recommends a staged rollout with event review, override testing, and clear ownership of exceptions.
If the workflow is sensitive, configure for caution first and autonomy later.
That is usually the correct operational sequence.