Db2 for z/OS - The AI Database



New Support chatbot for Db2 for z/OS: Six things to know

By Ann Hernandez posted Mon January 20, 2020 05:48 PM

By Ann Hernandez and Shu Li Kragness.

The Db2 for z/OS Support community was recently enhanced with the addition of the Db2 for z/OS Support chatbot, which is a self-service question-and-answer tool that could save you time. This blog post has information that will help you get the most out of the chatbot when you have questions about Db2 for z/OS but might not want to open a case.

1) How the chatbot works

The chatbot uses Watson Natural Language Understanding and machine learning technologies to provide answers to your questions about Db2 for z/OS. The chatbot is not a search engine; rather, it uses natural language technology to find the best answers to your questions.

To use the chatbot, click Chat with Support.  The chatbot assesses what you typed and determines your intent, which is similar to the "theme" of what you type. For example, if you type "how do I move to Db2 12?," the chatbot can assume that you are asking about migration. The chatbot then tries to find the best answers within the available migration information and displays one or more answers in the chat interface. The chatbot always asks you if the information it provided was helpful, and your YES or NO response can help the chatbot learn and get better at answering questions, over time.

2) What information does the chatbot have access to when answering user questions?

The Db2 for z/OS Support chatbot has access to hundreds of thousands of publicly available documents, including IBM Knowledge Center topics, IBM web pages, technotes, articles, APARs, and even some information that is outside the ibm.com domain. The fact that the chatbot has access to so much information is great, but it means that the chatbot often needs feedback from users so that it can successfully answer the questions. 

Here's an example of how user interactions can help the chatbot learn.  Before enabling the chatbot, we did a lot of internal testing. One test case was: "what is the default filter factor?"  As you might expect, the chatbot didn't know that this was a performance-related concept and therefore provided answers that weren't very helpful. We persisted in the chat by indicating that the answers weren't helpful and by specifying which category (from a list) was appropriate. Eventually, after several back-and-forth interactions, the chatbot provided the content that we hoped to find, and we let the chatbot know that this information was helpful. The next day, we repeated the test and the chatbot provided the right answer on the first try! This demonstrates how interacting with the chatbot can help it learn, which will--over time--increase the number of correct answers to future questions from you and other users.

3) What kind of questions can the chatbot answer?

The chatbot on the Db2 Support Community page has the potential to answer many types of questions you might ask, but it can't answer all questions, as you'll see in the table below. 
 Green checkmark The chatbot can typically answer questions.... Red X  The chatbot cannot answer questions...
About Db2 for z/OS and related products, including Db2 AI for z/OS, Db2 Analytics Accelerator, and the Db2 Accessories Suite offerings. About other IBM products that you use.
Example"What is the latest version of Maas360?"
Many other products offer their own Support chatbots on their respective Support community pages.
About information that is publicly available (such as IBM Knowledge Center, technotes, articles, and so on). About information that is not publicly available
Example: "I work for ABC Company. What is my customer number?"
That use words that are clear and specific. That are vague or ambiguous.
Example: "I have an install problem."
If prompted, try to rephrase your question so that the chatbot might be able to find a good answer.
About how to troubleshoot known product problems, when documented solutions might be available.  About situations that you know require log review or troubleshooting expertise of a Support team member. The chatbot can lead you to open a case and work with someone on the Db2 Support team, or you can click Open a case instead of engaging with the chatbot.
About how to open a case or check the status of a case.  That request a live agent or human being. Db2 for z/OS does not offer live-agent support at this time.

4) What happens if my initial question doesn't generate a good response?

Don't give up! The chatbot is always learning and it needs your input to get smarter. Here are a few tips:
  • If the chatbot indicates that it is "still learning," try to modify your question. (For tips, see Question 5). 
  • If the chatbot provides answers that aren't helpful, answer NO to the question about whether this information helped. 
  • If the chatbot offers you a list of categories or subcategories, select the one that is best, or select "None of the above."
  • If you can afford a few minutes, let the chatbot try more than once to provide a good answer.  When it does provide a good answer, be sure to click YES to the question about whether the information helped.

5) How should I phrase my questions?

There are no strict rules, but here are some tips:
Because the chatbot is optimized to expect natural language, communicate with the chatbot as if you are talking directly to a colleague.  
  • Ask the question in words that make sense to you. Don't try to guess the words that make sense to IBM or to the chatbot.
  • Avoid pasting full messages, code snippets, log data, dump titles into the chat. Message numbers, abend codes, and SQL codes are fine, but use them with natural words, too. 
  • Avoid using acronyms and jargon unless their use is pervasive in the Db2 for z/OS product documentation and user community. 
When asking the chatbot about a problem situation, phrase your question so that the chatbot knows you are troubleshooting.
  • Include key symptoms in your chatbot question. Example: "I'm getting an abend 0C4 in DSNUCAML."
  • Avoid being too vague. In addition to your key symptoms, include words about what you were doing when you the problem occurred.  Example: "I ran RECOVER TABLESPACE but now I'm getting an abend 0C4 in DSNUCAML."
  • Unless you are asking about migration problems that are specific to a single version, omit the version of Db2 that you are using. If you get too many results, you can ask again with the version.
  • To troubleshoot a problem that you suspect is common and that might already have an available solution, include the word "troubleshooting" in your question so that the chatbot will know you are troubleshooting and not just trying to look up a specific piece of reference information.

If the chatbot still can't find a good answer, the chatbot might ask if you want to open a case. If you click YES, the transcript of your interaction with the chatbot will be passed to the Support team member who works on the case. 

6) Can any human beings see my conversation with the chatbot?

Yes! A human being reviews the monthly chatbot logs. The chatbot focal point takes appropriate actions to make new content available to the chatbot or to make existing content easier for the chatbot to find in the future.  Also, you can type "Feedback" in the chat window to provide your feedback on your experience with the chatbot, and your feedback is included in the log that is reviewed.  

Thank you for reading about the Db2 for z/OS chatbot. We hope you'll try it and be patient as the chatbot continues to learn. Please let other Db2 for z/OS users in your organization know about this blog post so they can benefit from the chatbot, too.
1 comment



Mon January 27, 2020 12:52 PM

Great info - love the "CAN" and "CANNOT" table - very helpful.