Our summary
Changing an existing chart can still make an AI assistant repeat a slow planning process. João Viana describes Lightdash’s Fast mode, which uses Jev to identify small follow-up edits. The aim is to change what is already there without starting another full assistant run.
Jev chooses an edit type and the needed fields or values from known options. Lightdash’s software prepares the change, checks confidence levels, and asks Jev whether the plan covers the whole request. If these checks fail, the normal assistant takes over; otherwise, the software applies the edit.
Lightdash reports median replies of 1.7 seconds instead of 17.5 seconds across 25 scripted follow-ups in four conversations. That small comparison does not establish general accuracy. Testing also found missed filter values and broken handoffs, showing why teams need to check the real user route and fallback behavior.
Key takeaways
- Choose from chart fields, values and edit types already available; ordinary software applies the selected change.
- Check both the confidence of each choice and whether the planned edit covers the whole request before using the shortcut.
- Test the route customers actually use and track how often the shortcut runs, not just whether it reports errors.