Product setup and connection
Use one current Phoenix Tree AI product, connect the available read path and configure it for the team's own records and review routine.
Custom builds
Custom work starts close to current product work already underway. The best project is narrow enough to test, connected to a source we can inspect and important enough that a person already checks it by hand.
Ways to work together
Use one current Phoenix Tree AI product, connect the available read path and configure it for the team's own records and review routine.
Extend a current product around one missing source, calculation, approval step or output that can be tested against real work.
For a repeated workflow outside finance or commerce, first prove that the data, rules and acceptance checks fit the same evidence-led system.
Current bases: Global Markets Research and Open Commerce Workbench. Both remain in development, so a custom engagement does not turn either one into a supported public release.
A workable brief
Not a fit today
Process
Start with the repeated task, current operator, source system and one example of a good result.
List every source, deterministic rule, model judgement, review point and external permission.
Agree what will be built, what will stay out and how the finished path will be accepted.
Test the narrow path against representative inputs, including missing and conflicting records.
Deliver the files, configuration, connection notes and operating instructions in the client's own environment.
Handover rule
Price and delivery time are quoted only after the evidence map and acceptance boundary are clear. There is no public package price while the offer is still being defined.
First email
Three lines are enough: what repeats, where the source data sits and how a person decides that the result is correct.