Model : custom form processing model
Current license : O365 E3 license
Current phase of project : POC . There are no downstream systems attached. in this POC , We simply extract the data and show How AI builder can extract data in best way.
Next phase : moving to pod , which means , the custom model and the power automate flow which is using the model, will
process Prod documents and data will flow to downstream production systems.
Below are my questions about the AI model and capacity w.r.t licensing
- Can I Share and edit the Model with others so that others can train the model ? But I should be able to only publish is it possible ?
- Currently the model is created by me but I think while we move to production we need to make the model owned by a service account , I think this is the right design so that the ownership doesn’t lies with individual , is it transferable ?
- Currently I have created the flow to use the model. This flow will take documents from one drive ( my one drive folder) and I trigger it manually for testing purpose. If we move to prod , we need to create the flow from service account , is that the right design?
- We are now extracting data form different documents , is there any limitation that how many document I can process using this custom model ? Given the license I mentioned above?
- I processed 25 documents at a time once , only 20 documents were processed . Is there any limitation that my power automate flow using the custom model can process only "N documents at a time" ? May be limitation for license
- My final test would be against 300 files and I will be presenting the data points for testing to business, can I do that with my current licensee ? Some of the documents are more than 1 pages ,it may be 2 or 3 pages also . So totally it will be close to 500 pages.
- If upgrading license is mandatory for point 1 and 6 , Then I need to know that in advance let me know experts , please.
Below are my questions about the Feedback loop of the AI model and capacity w.r.t licensing
- Can I create the feedback loop with my current license ? Am not just talking about feedback loop in the PA flow . I think feedback loop involves routing documents to be processed to Microsoft cloud and Dataverse so with my current License , is it possible to do it ?
- Let's say the routing to feedback loop has to happen on confidence score of "one field " , but I might have got all other fields correctly extracted from the same documents. In other words, I am training the document for only one particular field. How do we convey this information via feedback loop so that the trainer knows this ?
- Also , the Training on the Feedbacked documents will happen only on agreed frequency , say weekly or monthly. so there is a possibility that similar kinds of document which is lacking confidence on same field will be routed again .
say eg : 25 docs of type A for the feild F1 wiould have got routed in a month's time. but there is no need to Label all 25 documents on F1 rather 10 is enough . how to pass on this information to feedback loop or data verse ? this question is similar to above question Q2.
@Antrod @gbego @joef