top of page

IBM Slack Integrated ChatBot

User research  | Conversational UX

IBM Watson based chatbot integrated with slack.

Context

When organization receives any new project , bid or proposal it will go through many team's approval.

 

Quote creation for a bid, project proposal may be challenging for many new employees and they may take help of senior employees to clarify their doubts. And it may also be challenging and frustrating for senior pricing executive to clarify 20 to 30 people’s doubts everyday.

To address this problem “IBM”  developed chat bot which answers the questions or doubts raised by

- Q2C team

- Sales team

- Pricing team

Problem

problem_edited.png

Organization's Geo pricing manager gets nearly 15 to 20 calls from the Q2C , sales and pricing teams to clarify their doubts.

Solution

solution_edited.png

Diverting those 15 to 20 calls to bot so that employees get their doubts clarified as initial help.

Impact

- 60 to 70 % Reduction in daily calls to Geo manager

- 50% of pricing, sales and Q2C team's time saved

Project goals

- Facilitate a dialog communication between bot agent and employees

- Solving the doubts of employees that arises while they quote a proposal or bid.

- Delivering the experience via a familiar , understandable conversational flow

Team

2 Designers, 1 Product owner
1 backend and 1 front developer

My Role

Research, analysis, insights, persona,

user journey map, dialog flow use cases

Duration

3 Weeks.

June 2020

Design approach

Timelines

01- Discovery and Planning

Since project had very limited time we planned and sent questionnaire to following teams.

Few sample questions from questionnaire.

To stakeholders

  1.  Who is your target audience?

  2.  What are the process for pricing that you follow?

  3.  What are the most frequently asked questions?

  4. What are the existing platforms your audience refer to?

  5. Where is the data coming from?

  6. How much time do you spend in day to address the queries?

  7. What are your business goals?

To technical team

  1.  Is there any front end framework that will be used?

  2. In which programming language AI logic will be written?

  3. Is there any API will be reused?

  4. How development environment is set?

  5. Anything that is platform dependent?

Project Documents

Research Plan Doc.jpg

Questionnaire to all users

- Q2C team (8 participants)

- Finance team (4 participants)

- Sales team (4 participants)

- Pricing team (2 participants)

Mural tool screenshot

persona 1.png

Insights

1. Seamless conversation - The bot should be able to recognize and carry out conversations with cross-function teams

2. Origin of Data - Suggest options (structured responses) as effective inputs that will help the user get what they are looking for or trying to do. Rolling back the user to the publisher may not make the bot very helpful.

3. Tone of the bot - The bot should mirror the audience demographics and the care situations they’re likely to find themselves in.

4. Analytical - A great chatbot will be able to suggest and recommend products and services for a user based on current or previous interactions.

5. Divergence - Guiding the users to the right human agents or resources in cases of cross-function approvals

6. Accessibility - Bot should be able to read text aloud for the visually impaired, for example.

02- Define

Personas

stakholder Persona.png

Defining Bot.

  1. An intelligent platform to resolve queries around the pricing structure of deals in Organization “X” which is not available on organizations guiding site.

  2. It will  reduce the load of callers on the Geo Pricing Manager and takes the traffic.

  3. Bot should be versatile and help contain the last minute pressures

  4. Maintain consistency and transparency in the system

References/ Benchmarks

Siri by Apple

The software is adaptable to
users individual language usages,
Searches, and preferences, with
Continuing use.

Haptik IVA

Haptik builds intelligent Virtual
Assistant solutions that enhance customer experience and drives ROI for large brands

Mitsuku

Mitsuku replies to your question in the most humane way and understands your mood with the language you’re using.

03- Design

Customer Journey Map

customer_journey_map.jpg

Conversation Design Principles For Chatbots

• Introduce your bot - Be crisp and clear about the identity of your bot and set the right expectations.
• Orient the user / Recipient Design - Tailor the utterances so the user can understand
• Be proactive - Actively identify scenarios where guidance might be needed for users that don’t know what to ask or how to ask it.
• Decrease ambiguity / Minimisation - Summarise complex input from the user without sacrificing understanding
• Provide clear actions - Provide a menu of structured responses narrow the scope of the input.
• Guide the user to their goal - Ensure successful interactions by providing guidance and feedback during complex processes.
• Repair - Redo the utterance, if users encounter trouble.
• Get feedback - Collect user feedback to improve your bot’s relevance and accuracy over time.

Shape Of The Conversation (Conversational UX)

How many utterances are also sentences?
Multiple

Who talks the most?
The advisor talks most in the entire conversation.

How often do they speak at the same time?
Twice or Thrice they speak at the same time.

What is the longest silence between them?
The pause when the answer has been conveyed and the user is trying to understanding and looking for any other query.

On what line(s) do they identify the topic of the conversation?
When the conversation starts and the question comes up, the topic is identified.

Dialog Flow/ Architecture Structure

Summons - Open a conversation “hello”
Capability Check - To start to the agent about what it can do or about the scope of the application.
Repeat Request - To get a repeat of the prior utterance.
Paraphrase Request - To get a paraphrase of the prior utterance
Continuer - The type of utterance the user can do, which bids the agent to continue talking
Sequence Closer - To close a sequence, such as an inquiry or request, that the user opened.
Sequence Abort - To cancel a sequence before completion
Conversation Closer - To close the conversation.

Dialog Flow/ Architecture

Use Case Senarios

Scenario 1: First time user

Scenario 2: Normal question and answer. In this case human gets answer

Scenario 3: When bot don’t have answer

Scenario 4: When human may not able to convey message (No right utterances)

Scenario 5: When human may not able to convey message and even bot don’t understand

Scenario 6: When human make spelling mistake

Error Case Scenarios

Scenario 1: Spelling mistake

Scenario 2: No internet

Scenario 3: Back end server problem

Scenario 4: Poor connection

Scenario 5: Loading

Wireframes

Testing

Project did not have time and budget to allow designer to conduct testing. If time and budget given to designer then I would have conducted "feedback surveys" to analyse  the patterns in "quantitative" and "qualitative" data by including multiple choice questions and also by including open ended questions.

But stakeholder and development team together collected user feedback, whether chat-bot is able to resolve their issues or no.

Final Outcome

It's been more than a year now this app is launched and utilized by more than 100 end-users globally.

​

As a impact there is

  •  60 to 70 % Reduction in daily calls to Geo manager

  •  50% of pricer, sales and Q2C team's time is getting  saved.

Watty_Mockup_Screen.png

To comply with my non-disclosure agreement, I have omitted and obfuscated confidential information in this case study. 

App icon and name are used only for representation in this case study.

bottom of page