If you are unable to create a new account, please email support@bspsoftware.com

 

News:

MetaManager - Administrative Tools for IBM Cognos
Pricing starting at $2,100
Download Now    Learn More

Main Menu

Cognos Analytics Assistant

Started by dougp, 15 Aug 2025 04:39:07 PM

Previous topic - Next topic

dougp

One of the new/improved features of Cognos Analytics 12.0.4 is the AI Assistant.  And I know that my management team is chomping at the bit to use any new AI feature that becomes available.

So I thought I'd try using these features.  Blocked!  "This data source has not been enriched."

Looking at the manual option for enriching a package, it looks like all I can do is tell Cognos which columns are date/time or location.

What I was hoping for is that my users can ask natural language questions of packages containing thousands of columns of non-trivial data.  So a user might ask a simple question like, "Show me expenditures by month for the last biennium for job number XE343401."

So, Cognos would need to know:
  • A job number is WorkOrderGroupNumber or WorkOrderNum + GroupNumber.
  • Expenditures could be called Expeses, Expended, Actual, etc.
  • "last" biennium is 2023-2025.  The 2023-2025 biennium is the same as the biennium with begin year 2023 or the sum of FY2024 and FY2025.  Also relevant, for system load and query performance reasons, is that since the user is asking for only one biennium of financial data, using the biennium begin year column on the fact table will help because that is how the database is partitioned.  You get a query performance improvement of 2x to 40x if you craft the SQL properly.
  • There are 24, 25, or 26 "months" in a biennium, depending on whether the employee works primarily with budget, program management, or accounting subjects.
...and when the user is asking for months from financial data, they are not asking for something that is really date/time data.  They are fiscal periods, not really months.  But they still need to be presented in the correct order, and there may be a need to forecast even though it is not considered time-based data (since only accountants - and not the Cognos devs - think there are 13 months in a year).

I do have available to me a data catalog that maps business terms to data objects.  So that could help.

How do I feed synonyms and other relevant vocabulary and business relationships to the AI?