Build an enhanced IT help desk chatbot on IBM i with Watson Assistant

Deploy a chatbot prototype that can understand and automatically solve
password-related issues on IBM i

Watson Conversation is now Watson Assistant. Although some illustrations in this tutorial may show the service as Watson Conversation, the steps and processes will still work.

Did you know that as per recent studies, an average of 20% to 50% of all help desk
calls are password related? Mundane tasks can be a headache for help desk teams,
especially when higher priority jobs need attention. Virtual assistants (also known
as chatbot) for IT help desk can automatically process tickets that do not
need deep expertise such as managing access credentials and password-related issues,
allowing your staff to focus on critical issues and value-added tasks. For instance,
a chatbot can analyze the root cause and automatically solve password-related
issues.

In this article, you will implement a simple chatbot to illustrate this kind of
solution.

It can understand user profile-related issues, and, if possible, automatically solve
the problem. As the icing on the cake, it can also answer to
queries about system utilization, such as processor utilization or the percentage of
system auxiliary storage pool (ASP) used.

This chatbot links the IBM® Watson® Assistant service and IBM i (back
end) to Slack (front end), which is the messaging team collaboration tool. It will
be implemented in the prototyping mode with Node_RED, and will run on IBM i.

You are now going to instantiate and configure the required services, import the
Watson Assistant workspace (predefined dialog training) and Node-RED flow
(chatbot back-end program), and test it!

What you’ll
need

To implement this chatbot on your IBM i system, you need:

• An IBM i 7.3 partition with Node-RED installed (if you need to install Node-RED
  on your IBM i, you can refer to my previous article Running Node-RED on IBM i: Installation and first flow,
  developerWorks, October 2017).
• An IBM Cloud (earlier known as IBM Bluemix®) account to be able to
  provision Watson Assistant services.
  Register for your free trial
  account or log
  in to IBM Cloud  if you already have an account.
• A Twilio account (trial version available), and a Slack account (free version
  available).

Architecture
overview

To implement this IT help desk chatbot on IBM i, four components must be deployed and
configured. Figure 1 shows the components deployed to support the chatbot solution
implemented in this article.

Figure 1. Chatbot architecture overview diagram

The four components are:

1. IBM Watson Assistant service providing natural language capabilities:
   intents/entities recognition and dialog management.
2. Slack software used as the user interface, in which you will create a Slack bot,
   a component required to interface Slack with your own chatbot program (#3).
3. Your chatbot back-end program, implemented with Node-RED on your IBM i system.
   It is the core of the solution, called application logic, linking all
   the components: Watson Assistant service (#1), user interface (#2), and IBM i
   system and database (#4).
4. IBM i system and the IBM Db2® database for retrieving and updating IBM i
   user profile (*USRPRF) attributes.
   In this prototype, the Node-RED program
   can directly query the database or run the program (integrated in #3). But, in a
   multi-system configuration, this component must be independent and deployed on
   all systems. Then data and services are available through web services
   [Representational State Transfer (REST) APIs].

In this article, let us see how to set up these components.

IBM Cloud: Deploy and train Watson Assistant service

In this first part, you will instantiate, train, and test Watson Assistant, (the
Watson cognitive service in charge of dialog with a chatbot user).

The IBM Watson Assistant service combines machine learning, natural language
understanding, and integrated dialog tools to create conversation flows between your
applications and your users.

Perform the following steps to create your own Watson Assistant instance:

1. Log in to IBM
   Cloud using your IBM Cloud account.
2. Select the region and space you want to work in.
3. Click Catalog, then click Watson in the left
   navigation panel. The catalog displays the available Watson services. Click
   Assistant to view the service configuration page.

   Figure 2. IBM Cloud platform, Watson services
   catalog

   

   
4. Provide a unique service name (or use the default service name), and choose plan
   Lite (free). This plan allows 10,000 API calls per month,
   five workspaces, 25 intents and 25 entities. If you need more, you can subscribe
   to Standard or Premium plans. Click Create.

   Figure 3. Assistant instance creation

   

   
5. In the new service dashboard that is displayed, click Launch
   tool to launch the web tooling interface where you can define
   intents, entities, and the dialog for conversation. A new page opens in your
   browser. Click Log with your IBMid.

1. Import the iHelpdesk.json workspace by clicking the
   Import icon (highlighted on Figure 4).

   Figure 4. Watson Assistant tool

   

   
2. Click Choose a file, and select the iHelpdesk.json file,
   previously downloaded from “Downloadable
   resources“.
   Then, click Import.

   Figure 5. Importing a workspace

   

   
3. A new workspace is created and opened. To view the conversation dialog, click
   the Dialog tab.

   Figure 6. Watson Assistant workspace

   

   

Watson Assistant: Understanding a conversation

You can have a look at the workspace content to get an idea of the questions you
could ask to the chatbot (in the Intents and
Entities tabs) or understand the way the dialog flow has been created
(in the Dialog tab).

For example, you can try to understand the Reset Password flow
implementation in the Watson Assistant service. Figure 7 illustrates how the
#lost password intent dialog has been designed for this demo,
where:

• Text within blue boxes are messages sent to a user
• Text within green boxes are actions identified by the Watson Assistant service that
  the chatbot program will have to run
• Orange squares are conditional statements
• Green circle is dialog entry point for this intent
• Red circles are dialog exit points for this intent

Figure 7. Reset password dialog flow

The chatbot program on IBM i requires the workspace ID and credentials to connect to
the Watson Assistant API. To retrieve these values, click the
Deploy icon and then click Credentials
(highlighted in Figure 8).
Copy the workspace ID, user name, and password.

Figure 8. Watson Assistant credentials

Your Watson Assistant workspace is now configured. Intends, entities, and dialogs
are ready to be used by your IBM i chatbot program.

If you want to learn more about Watson Assistant tool and interface, refer to the
Watson Assistant documentation website.

Now, let’s configure Slack, a software as a service (SaaS) collaborative solution
that will provide the user interface of this chatbot.

Slack

Slack is a cloud-based set of team collaboration tools and
services.

Slack works like an Internet RelayChat (IRC), organized in channels
corresponding to many topics of discussion. The platform also makes it possible to
keep track of all exchanges, allows file sharing within conversations, and
integrates external services (such as GitHub, Box, Skype and so on) or your
chatbot implemented with Watson Assistant. To integrate your chatbot, you need to
use Slack bot, an integration solution that allows to communicate with your chatbot
program which in turn communicates with Watson Assistant.

First create a Slack workspace or use an existing one if you have sufficient
authorization. If needed, refer to Slack’s how-to on creating a new workspace. During
this process, you must enter a name that will identify your Slack workspace and
define the workspace host name (by default, <WORKSPACE>.slack.com).

In this article, we will use ihelpdesk as the workspace name, and
ihelpdesk.slack.com as the host name (don’t forget to use your
own details).

Open a new browser page on this host name to get access to your Slack workspace.

Figure 9. Slack Workspace on ihelpdesk.slack.com

Perform the following steps to create a custom bot in your Slack workspace:

1. Open a new page in your browser and navigate to the Slack Custom
   Integration-Bots creation page using the following URL:
   https://<WORKSPACE>.slack.com/services/new/bot (replace
   WORKSPACE by your own workspace name).

   Figure 10. Slack bot creation interface

   

   
2. Name the bot (you could change name after, if needed) and click Add bot
   integration. In this article, we will use ibot as
   the bot name.
3. A new page opens containing the API token (the bot’s secret token).
   Token
   is used to connect your chatbot application (running on Node-RED on IBM i) to
   Slack servers by using a custom TCP/IP network API called Slack Real Time
   Messaging (RTM) API. This allows all user interactions with the Slack bot to be
   sent to your connected chatbot application.

   Figure 11. API Token in Slack bot setting page

   

   

   Copy this token because you will need it to configure Slack nodes in the
   Node-RED flow.

4. Switch back to the Slack workspace page.
   You can see a new app, the Slack
   bot (highlighted on Figure 12). It is currently not connected. It will appear as
   connected after creating the chatbot program with Node-RED (refer to the “Node-RED” section of this article).

   Figure 12. Slack new app available

   

   

   Your Slack
   workspace is configured. The Slack bot application (a virtual user) is now ready
   to be connected to the Node-RED application you will create.

Twilio

Twilio is a cloud
communications platform as a service (PaaS) that allows developers to
programmatically integrate voice, messaging, and VoIP into web and mobile apps.

Create a
trial account (or connect to Twilio using your existing account) and obtain a SMS-capable
phone number.

Check the following references for detailed information about Twilio
configuration:

1. Sign up for a Twilio account
2. Get an SMS capable phone number used to send text
   messages.

After completing the configuration, retrieve and copy the following information that
will be required to configure the Node-RED Twilio node:

1. Account security ID (SID) and authentication token from the Dashboard page.
   

   Figure 13. Twilio account SID and authentication
   token

   

   
2. An SMS-capable phone number (see Figure 14). To access this, click All
   Products & Services on the home page, and then click #
   Phone Numbers.

   Figure 14. Twilio active numbers

   

   

   Note: If you
   want to send text messages to another phone number, you must declare it on
   the Verified Caller IDs page.

   Figure 15. Twilio callers

   

   

   Your Twilio service
   is configured and ready to be used for authentication (2FA).

As all cloud external services are ready (Watson Assistant, Slack, and Twilio),
you can now create the chatbot program with Node-RED on your IBM i system.

Node-RED

Node-RED is a flow-based
programming tool, originally developed by the IBM Emerging Technology Services
team (in early 2013) and now a part of JS Foundation.

Get the code.
To get the code, download the
Node-Red_chatbot_flow.v1.1.json file from the “Downloadable
resources” section of this article, and copy the entire file’s content into
clipboard.

Import the code.
Go back to the IBM i Node-RED editor in your
browser, click the menu at the upper-right corner and click Import
-> Clipboard. Then, paste
the contents of the Node-Red_chatbot_flow.v1.1.json
file content that you
copied. Click New flow, and then click Import.

Figure 19. Import chatbot flow

Click somewhere on the blank sheet to paste nodes.

Have a quick look at chatbot flow. Figure 20 highlights the main logic of the
flow.

Figure 20. Chatbot flow overview

The numbering in the figure corresponds to the numbering in the following procedure
that explains the flow.

1. Get the text entered by the user on Slack.
2. Send it to the Watson Assistant service. This service understands the intent,
   identifies entities, defines what to answer.
   It sends back an answer for the
   Slack user and initializes variables in the context part of the JSON object
   exchanged. The value of context.Action variable indicates whether
   an action has to be executed by the chatbot (and what action) or not.
3. Depending on the Action variable value, the switch case node will
   route to perform one or two steps (from within step 4 to step 8).
4. Answer directly to the Slack user.
5. Retrieve resource usage value from IBM i (by querying Db2 for i services).
6. Generate a temporary password and an authentication code, and send text.
   If
   you don’t want to use SMS to send code, you can configure the email
   node to send it by email.
7. Change the user profile (*USRPRF) IBM i object. First test if password reset is
   allowed (password not *NONE), and then run the system CHGUSRPRF
   shell command.
8. Retrieve the user profile parameters from IBM i (by querying Db2 for i services)
   and determine the possible cause of the connection issue.

After performing steps 5, 7 or 8, the result is sent to Watson Assistant to
transcript in a natural language sentence.

Link nodes (#a, #b) allow to make a connection back to the beginning of
the flows (from a to b), without overwriting the flow representation.
Link nodes allow to set up a mechanism like the symbolic links of
UNIX file systems.

Update the code.
You must now configure nodes corresponding to
cloud external services previously created. Figure 21 shows the nodes to be
configured.

Figure 21. nodes to be configured

1. Watson Assistant
   Double-click the Conversation-Slack node
   (Figure 21, #1), and enter the user name,
   password and workspace ID (captured in the
   “Watson Assistant: Understanding a
   conversation” section of this article.
2. Slack
   Two Slack nodes must be configured: Slack bot in and
   Slack bot out, and they both have the same name:
   helpdeskbot.
   Double-click each node (Figure 21, #2), and
   enter the bot API token (captured in the “Slack”
   section of this article).
3. Twilio
   To configure the Twilio node (refer Figure 21, #3a),
   double-click it, and select the External service value for the
   Service field.

   Figure 22. Twilio node configuration

   

   

   Then,
   click the Edit button (highlighted in Figure 22) to add a
   new Twilio API (twilio-api).

   Figure 23. Twilio node account configuration

   

   

   Enter
   the required values in the Account SID, From, Token and
   Name fields (see Figure 23). You should have captured
   those values in the “Twilio” section of this
   article.

   Then, configure the function node (Figure 21, #3b), named
   Build SMS.
   This node build message to be sent by text,
   and also defined the target phone number (this is only a simple demo, but
   think about dynamically retrieving this phone number from an enterprise user
   directory, for instance, Lightweight Directory Access Protocol (LDAP) or IBM
   Directory Server).

   Initialize the msg.topic
   variable with the phone number to which you want to send messages. Remember
   to replace xxxxxx with this phone number in the code (see Figure 24).

   Figure 24. Twilio function node

   

   

Deploy the code.
Click Deploy to deploy and
make your application live.
A message, Successfully deployed,
appears at the top of the window.

You can now test the chatbot.

Test chatbot

Go back to the Slack window, and see that the virtual user is now connected (as shown
in Figure 25).

Figure 25. Slack bot connected

This chatbot was trained to answer the following four user intents:

These are just examples of resolving password-related issues. You can find many ways
to improve it by changing the Watson Assistant dialog flow or the Node-RED
flow.

Remember that the quality of chatbot is mainly related to the quality of the Watson
Assistant training. And, as you could see in the “Watson
Assistant: Import a conversation” section, the current training (number
of sample sentences by intent) is not sufficient. You can add more sample sentences
to train Watson Assistant better.

So, now, it is your turn.

Switch back to the Slack interface.

Click the iBot app on the left pane, and enter your questions.

For exhaustive tests, don’t forget to change the parameters of test user profile
(*USRPRF on IBM i), for example, password *NONE or expired, user profile *DISABLED
due to too many wrong passwords entered, and so on.

Conclusion

In this article, you could see that it is quite easy to create a chatbot to
automatically support simple ticket resolution tasks.

This easiness is due to three main factors:

• Watson Assistant in IBM Cloud, providing a powerful tool to design and
  support dialogs in a natural language
• Node-RED on IBM i, a Node.js development application allowing quick and easy
  prototyping
• Db2 for i services, simplifying access to many pieces of IBM i system
  information through a database interface rather than complex commands or
  APIs.

So, next, what will you request to your chatbot to do? Solve printer issues, the
second source of tickets in IT help desk organizations? Provide FAQ interface to
your IBM i users?

There are so many ideas and their implementations are now possible on your IBM i.
Let’s do it!

IBM Watson Assistant

Node-RED

DB2 for i services