How a financial or treasurer report is reviewed
Code checks the figures first
Before a model writes the report, code checks the figures.
One model drafts
A first model drafts the financial or treasurer report.
A different model family challenges it
A model from a different family reads the draft and challenges it.
A last look for what doesn't fit
A third model looks for anything that still doesn't fit.
Disagreements go to a person
If the models still disagree, that point is flagged for a person to review.
The report shows which models checked it
The report keeps a record of which models drafted it, challenged it, and did the last check.
People approve the report. Steward never moves money. Q-Think is a review your treasurer and your board can read. It does not decide for them.
Everyday questions stay on one model
A simple request stays with one fast model. If that model is unsure, a confidence check sends it on for a closer look.
Also part of Q-Think
Steward keeps answering
If one AI provider is unavailable, Steward switches to another and keeps answering.
Thumbs up or thumbs down
You can mark an AI answer thumbs up or thumbs down.
Coming later
These are not part of Q-Think yet.
Legal answers
A review of legal answers is coming later.
Meeting minutes
A cross-check on meeting minutes is coming later.
Executive sessions
Splitting an executive session across models, so no single AI provider sees the whole session, is coming later.
Questions
- Financial and treasurer reports. Code checks the figures first. One model drafts the report, a model from a different family challenges the draft, and a third looks for anything that still doesn't fit. If the models still disagree, that point is flagged for a person. The report keeps a record of which models checked it. People still approve the report, and Steward never moves money.
- A simple request stays with one fast model. If that model is unsure, a confidence check sends it on for a closer look.
- Steward switches to another model and keeps answering.
- Yes. You can mark an AI answer thumbs up or thumbs down.
- Not yet. A review of legal answers, a cross-check on meeting minutes, and splitting an executive session across models so no single AI provider sees the whole session are coming later.