Business Problem:
Currently, when the platform updates its underlying AI models, all users are forced onto the new version immediately. In a professional production environment, even subtle changes in how a model interprets instructions can lead to unpredictable behavior and output regression.
For users who have spent significant time "Prompt Engineering" their AI for specific, high-precision tasks (such as intent routing, data extraction, or strict formatting), these forced updates often break existing logic. A model that was previously reliable may suddenly change its tone, formatting, or compliance with negative constraints, forcing teams to manually re-test and re-adjust their entire automation setup.
Desired Outcome:
  1. Model Version Selection: Provide a dropdown menu within the AI Agent/Workflow settings that allows users to select specific model versions (e.g., GPT 5.2 vs. latest).
  2. Version Locking: Allow users to "Lock" their AI to a specific version so it does not update automatically when the platform releases a new default.
  3. Legacy Support: Ensure that older versions remain available for a reasonable grace period, giving businesses time to migrate and test their instructions before the old version is sunset.
Impact on AI Results:
Forced updates compromise the reliability of AI-driven automation. Without version control, we cannot guarantee a consistent experience for our customers. Providing versioning options will allow professional users to build stable, mission-critical automations without the fear of their logic breaking overnight due to platform-wide updates.