AI overkill

[Bob Cergol]

AI is an incredibly wonderful, powerful -- and "sexy" feature.

However I would like to be able to tell it to not look at certain fields.

Using the AI feature on an order report based on my core order table view AI generates 21 different reports -- 12 of which are using meaningless fields such as 40 character description lines that are not part of any index. Some others are numeric fields that store data irrelevant for management reporting such as revision number.

Being able to create an "include" and/or "exclude" list of columns that AI should consider would be a nice feature. Absent any entries it could behave as it currently does.

Or maybe this already exists and I just haven't found it yet.

Bob

[Carsten Sørensen]

Your suggestions are absolutely valid - however I have the following comments (explanation to the current AI behavior):

1. AI isn't looking at the indexes as the idea is to present also sub-reports that are actually relevant but not necessarily planned to be.

2. Very long fields are / should already be avoided - however even description fields are sometimes relevant as they are being used to store information that were maybe not taken into account when the table structure was designed. I.e. many times end-users find a way to create additional dimensions on data by entering data in a structured form into description type fields. Thats the explanation why they are currently being consider by AI.

3. The criteria for being displayed on the AI dashboard is that AI finds a minor number of distinct values for that column. It will however often pick irrelevant columns - as you point out and I agree.

To solve the issue the AI dashboard was / is intended to be used the following way:
A- Run AI
B- Remove irrelevant suggestions by closing the individual irrelevant reports (clicking x).
C- Eventually re-order (by dragging on the titles)
D- Eventually merge reports and eventually remove the underlying individual reports
E- Save the dashboard

By using above methodology you can prepare a dashboard with relevant sub-reports - and save it for management or whoever that needs it. Next time running the dashboard from the menu will refresh all sub-reports with the current data and you wont have to remove anything from the dashboard.

I can agree - and I guess that is what you point out - that it could be nice to prepare for AI use eventually using flags on the dictionary etc. - it might be a good idea especially if the same data-source is used with AI multiple times (which is often the case)!

Thanks for the feedback!

Best regards
Carsten

[Bob Cergol]

What you recommend makes perfect sense.

It also makes sense to me that "AI" needs to "err" on the side of giving more, not less, of what is potentially useful information.

That the question even comes up illustrates what an amazing feature it is!

What made me think of it was my frequent experience of showing end users some wonderful new functionality I've given them -- which automatically becomes the new baseline of their expectations -- and so they look past all the wonderful stuff and jump on peripheral issues. In the example I was writing about I was envisioning one of our managers clicking the button and then actually being distracted or confused by so many reports -- and then asking me: "Do I have to close them every time?"

It just occurred to me what a nice added feature for developers it would be to "help" AI based on their intimate knowledge of the underlying data.

Bob

P.S. Do you ever sleep?

[Carsten Sørensen]

Hi Bob,

Thanks for the input.

For the next release we have added a feature to mark columns that are AI N/A at dictionary level. Columns marked will not be considered for the sub-reports.

I realize that although it is easy to remove irrelvante sub-reports - at least in the situation where you apply AI to an existing report - for ad-hoc analyzing purposes - it could be helpful to be able to remove N/A columns - and in that case it can be done centrally using the dictionary instead of having to maintain each single report (see below).

Best regards,
Carsten

[Bob Cergol]

Does this mean, everything I ask for you will add to the product?
Bob