Get PeopleSoft Chatbots Production Ready in 4 Simple Steps

This post was provided by guest author, Logesh Balasubramaniam, Managing Director, SpearMC Consulting

PeopleSoft Chatbots using the Oracle Digital Assistant (ODA) platform offers a transformative, conversational user interface to help users accomplish a variety of tasks in natural language conversations. PeopleSoft customers can leverage this capability and reduce ongoing training and service desk costs by providing on-demand assistance to users where and when they need it. The process for PeopleSoft customers to get started with Chatbots is straight forward:

• Check if you meet the minimum requirements of PeopleSoft Application and Tools versions
• Purchase and set up an ODA tenancy (multiple options available, including free trial )
• Enable PeopleSoft to ODA integration
• Deploy and train delivered Chatbots for PeopleSoft

Note: The latest information on the above tasks can be found here - PeopleSoft Chatbots 

Chatbots and the more sophisticated Digital Assistants offer a personal, interactive, and engaging channel for users to communicate. It is therefore important for companies to personalize the delivered chatbots to establish a flavor that is unique to the company and its culture. Described below is the process to get the PeopleSoft delivered Chatbots personalized and production-ready so they can be an effective means of user interaction:

1. Personalize the Utterances, Intents, and Entities
2. Enable Q&A feature to surface Frequently Asked Questions
3. Customize the Dialog Flow to refine Bot Conversations
4. Leverage ODA Insight Analytics to optimize Bot Performance

Now let's look deeper into each of these steps...

1. Personalize the Utterances, Intents, and Entities

In Chatbot vocabulary, “Utterance” is anything the user asks, which is examined by Natural Language Processor (NLP) to extract the “Intents” - what the user wants and “Entities” – the data to support the intent. For instance, if the user asks, “What is the cost of my dental plan?”, the whole sentence is the utterance, “Plan Cost” is the intent and “dental” is the entity.

Each PeopleSoft Chatbot is delivered with a set of Intents (what it can do), Entities (what data it can take), and Utterances (What does it respond to). Oracle Digital Assistant provides different ways for customers to personalize each of these aspects.

Utterance Library Export/Import

Utterances associated with an intent can be easily exported to a CSV file for editing and the updated version can be uploaded back. Conversational behavior of users varies highly across geography, demography, industry, etc. So it is critical to train the Chatbots on how your users will interact with them.

Enhance the delivered list of utterances to train the chatbots on company-specific terminologies, regional variations, languages, etc. For Instance, PeopleSoft's utterance could be “Show my savings plan beneficiaries”.  However, your employees may ask “Show my savings plan nominees”, Chatbots need to be trained on such variations.

Utterance Quality Report

ODA Provides a quality check report on the Utterance library that highlights issues like Similar Utterances, Misclassified Utterances, etc. It is a handy report to run and correct these issues to keep the utterance library effective.

Auto-Complete Suggestions

Customers can add a list of auto-complete suggestions for each intent, so the Chatbot can show predictive texts when the user starts to frame their question.

Answer Intents

This is a special type of intent that customers can use to quickly answer user questions. These intents do not integrate with backend systems like PeopleSoft to pull data and can be used to respond with a static answer to questions. For example, questions like “What is IT Service Desk email?” can be directly answered by the bot without backend integrations.

Entities Export/Import

An Entity list associated with an intent can be easily exported to a CSV file for editing and the updated version can be uploaded back. PeopleSoft Chatbots are delivered with a comprehensive list of entities for each intent and customers will need to review and refine this list for their specific usage. For Instance, the entity list for “Benefit Plan Types” will contain all plan types supported by PeopleSoft, however, customers may be only using a few of the plan types, so the unused plan types can be removed.

2. Enable QnA feature to surface Frequently Asked Questions

The Question and Answer (QnA) feature in ODA enables the Chatbot to answer any general-interest questions by returning a set of question/answer pairs that closely relate to user input. It is a way for companies to use the Chatbot to surface the Frequently Asked Questions or other knowledge base documents relevant to user inquiries. The question and Answer library can be easily created by importing a CSV file, and it supports multiple languages. Advanced settings allow you to control the ranking of questions/answers surfaced to the users.

PeopleSoft customers can use the QnA feature to greatly expand the services offered by the Chatbot by building a comprehensive library of Frequently Asked Questions, not just limiting the bot to tasks performed in PeopleSoft but covering the entire business process. For instance, a Recruitment Chatbot can be enhanced with a comprehensive set of Questions and Answers that job candidates frequently submit to recruiters. A well-built QnA library can cut down the volume of support inquires handled by administrators, thereby lowering overall user support costs.

3. Customize the Dialog Flow to refine the Bot Conversations

The Dialog Flow definition is the model for the Chatbot conversations, one that choreographs the interaction between the bot and its users. ODA uses “OBotML” which is its own implementation of YAML to manage Dialog Flow. 

PeopleSoft Chatbots are delivered with a default Dialog Flow, however, consider this as a template that can be modified to suit different usage requirements. Listed below are few use cases for updating the Dialog Flow in a Chatbot,

• Controlling the default options prompted to users after answering a question.
• Linking an Intent to another related Intent. For example, when a user requests to show their leave balance, after displaying the leave balance, prompt for creating a new leave request.
• Review the usage insights to see where the conversation is dropping abruptly and update the Dialog Flow to end the conversation gracefully.

4. Leverage ODA Insight Analytics to optimize Bot Performance

The ODA platform offers detailed analytics on usage patterns of each Chatbot. These usage analytics offer insight on how users are interacting with the bot, its effectiveness in servicing user queries, rate of usage, timings, success/failure rates, etc. Some of the key features of usage analytics include the following:

• Conversation Metrics - Completed vs Incomplete, Average duration
• Intent Metrics – Fastest, Slowest, Average, Frequent Paths, Common Utterances, etc.
• QnA Metrics – Rate of questions/answers/categories, Clicks per question, Average answer count

The usage dashboard also offers a “Retrainer” feature that can be used to improve the intent resolution of the chatbots by training them using the live data that flows through the digital assistant.

For PeopleSoft customers, especially in the early stages of customizing and deploying the delivered chatbots, usage analytics can be quite powerful in evaluating the chatbot’s effectiveness with smaller pilot groups before launching to the entire audience. Once launched to the target audience, these analytics can be used to continuously improve the bot’s effectiveness as the usage requirements evolve over time.

Conclusion

Market research supports that Chatbots can be trained to answer around 80% of user questions and can save support costs by over 30%. Chatbots offer 24x7 support and with the proliferation of bots on the internet, enterprise application users have come to expect on-demand assistance rather than lengthy instruction manuals and training documents. PeopleSoft customers can now leverage the powerful Oracle Digital Assistant cloud service to quickly personalize and deploy the delivered chatbots for their application users.

This post was provided by guest author, Logesh Balasubramaniam, Managing Director, SpearMC Consulting

You can learn more from Logesh at his blog:  https://leanitdesigns.blog/