Honestly this release might be a little early for IBM as well. Any sort of verbose error message is at least a low risk information disclosure finding on a pentest and a verbose database error message is a strong indicator the application is vulnerable to SQL injection. Even if not vulnerable, it's blood in the water and should be fixed before hitting production.
[1]http://www-01.ibm.com/support/knowledgecenter/SSATW2_7.3.0/c...
I think that both IBM and Microsoft are doing a good job at "reinventing themselves", BTW.
edit: by 'clean' I really meant data that is in a regular structure with good example questions specified for each paragraph in input documents.
I driving an effort for a major bank to combine all sources of different data (traffic logs and external treat intelligence feeds - all of different formats) to create a comprehensive fraud alerting and security investigation system. And we got pretty impressive results utilizing Splunk as a data swiss army knife.
IBM is certainly capable of offering a competitive solution to this industry although they need to move faster.
WatsonAnalytics is a good start even at it's current "teaser preview" stage.
DB2 SQL Error: SQLCODE=-964, SQLSTATE=57011, SQLERRMC=null, DRIVER=4.16.53
Apparently it's very sensitive to data input formats and essentially it wants you to do the data babysitting work first by properly preparing and structuring the data before it'll do anything.
These are cool demos on "friendly" pre-baked spreadsheets but way early for any practical production deployments, until IBM will start supporting raw datasets.