A read-only statement of how much the system may do on its own with your bills. Field by field, it also shows the evidence that would have to exist before it may do more.
When to use it
When someone asks how far the automatic reading is trusted, or why a particular field is still being reviewed by hand.
Before you start
The AI document add-on has to be switched on for your workspace, and you need viewing rights on this screen. There is nothing to set up. Nobody has editing rights here, on purpose.
Where to find it
AI > Autonomy.
How it works
- The ladder has four rungs. At L0, a person reviews everything. At L1, the reading fills the entry and a person still posts it. At L2, the reading fills the entry and only the doubtful ones are put in front of a person. At L3, the reading posts within limits the owner sets.
- Purchase entries run at L0 today, so a person reviews everything.
- The table below shows, per field, how many answers have been judged, what share were accepted, how many are still pending, and what is still needed for the next rung. The bar is at least fifty judged answers, with at least ninety-eight per cent of them accepted.
- There is no switch on this screen, and there is not meant to be. A field moves up when the evidence says it has earned it, and the move itself is the owner's decision.
- You build the evidence by working normally. Every reviewed entry you save turns pending answers into accepted or corrected ones on the Accuracy Ledger, and this screen counts those.
Tips and common mistakes
- "Bar met" against a field means the evidence would support promoting it. It does not mean the field has been promoted, and nothing has changed in how it behaves.
- A field with few judged answers is not inaccurate. It has not been reviewed enough times for anyone to say either way.
Related guides
See also Accuracy Ledger, Exception Queue and Document Pipeline.