Building a Context Aware and Personalised Experience through behavioral triggers and tools

Improving Product Discovery and Retention for Einstein AI Chatbot

Role

Product Design & Research

Duration

5 Months, Aug '24 to Dec '24

Location

San Francisco, California
(Remote)

Client

As the Einstein Chatbot
transitioned to AI,

its significant shortcomings in conversation capabilities led to user annoyance and increased dropouts.

High Level Goal

Designing a human-centered AI chatbot experience for businesses using the Salesforce website with a specific focus on product discovery.

Months of UX Research

Comp Analysis, 10+Interviews & Contextual Inquiries, Mapping

3+

Collaborative Sessions

Hypothesis Validation, Research & Insights, Mentorship

15+

Major Critique Sessions

Presentation & feedback with 20+ UX & Industry Professionals

Over 2

Rounds of User Testing

Over 2 rounds of user testing and iterations of concepts

Multiple

I led a team of 8 Designers & Researchers to collaborate with Salesforce Team over the span of 4 months

which helped identify 4 key issues

Endless Scrolling of Chat

Repeated Push to Contact

Overwhelming amount of information forced users to scroll back & forth to find relevant details.

• Disrupted Flow

• Lost track of Conversations

Robotic Tone & Personality

Ineffective Product Discovery

and this was the result

Design Research, Process and Journey

1

Using ChatGpt - Simulated advanced chatbot capabilities by designing prompts to mimic Einstein, allowing for realistic testing of features.

Context awareness was a focal point

as users expected the bot to “remember interactions" and provide "tailored responses"

"It's like the bot forgets what I asked or it simply doesn't get it. And I don't know what to talk to it. It always sends me to Sales!!"

🎯

Competitive analysis and contextual inquiries -Amazon Rufus, ChatGpt and Einstein,
xxxx

Identified gaps in the bot's ability to engage in conversations and maintain a dialogue

Users struggled with when and how to start typing; scrollling back to track context.

Framing prompts that chatbot would understand was difficult, leading to irrelevant responses.

Felt overwhelmed by lenghty, unstructured chatbot responses, users abandoned the chats.

discovered recurring patterns by studying 10+ users

The project kicked off with a clear yet ambitious goal

To transform Einstein Assistant into a human-centered chatbot that delivers a personalized and context-aware experience.

This helped me identify key issues in the designed features — Suggestive Prompts, Clarifying Questions and Product Recommendations. Narrowing down on one use case—New Customers using chatbot to search products, helped me to target specific user flows.

Design Critiques with Salesforce team challenged me to think from behavioral perspectives and to prioritize shaping intuitive interfaces.

Advanced Prototyping and User Testing

To refine designs, I employed several methods

2

User Testing - Conducted tests with business professionals to validate features, focusing on Behavioral Triggers (Prompts for reframing or inactivity), Product comparison tools, Timeline usability (tracking conversation history)

Iteration and Prototyping x4
Informed by critique and User Testing at every stage, I designed mechanisms for various scenarios

Scenario 1

Clarifying Questions

Users hesitate to provide detailed input, expecting relevant responses.

Behavioral Trigger

Scenario 2

Feature Design

Users need assurance that their chat and content remains accessible while browsing. Chatbot redirection enables dynamic product searches across the website.

Chatbot Redirection

Scenario 3

Suggestive Prompts

Users expect a GenAI chatbot to understand their needs and be context-aware.

Behavioral Trigger

Scenario 4

Compare Products

Users expect simplified product discovery through quick info to decide.

Feature Design

Scenario 5

User Inactivity with Chatbot

Behavioral Trigger

Salesforce values customer retention and aims to maintain engagement.

More details!

Complete Case Study

Salesforce design team loved the unique approach and have adopted the technique in practice.

*PS - I faced a unique challenge. Einstein did not have the ability to continue a conversation, and I wanted to see how would Einstein talk if it was more capable. I proposed asking ChatGpt to mimic Salesforce Einstein Chatbot by designing a prompt. This method helped the team uncover truly valuable insights.

Reflections and Takeaways

Project management, leading a massive team and facilitating critiques and presentations


This project significantly enhanced my leadership abilities and elevated my project management expertise by consistently challenging me with strict deadlines, demanding deliverables, client satisfaction expectations, and the responsibility of ensuring product success.

🚀

Not hesitating to explore innovative means to achieve goals


Designing is never a defined process, infact it is a canvas guided by constraints, This project pushed me to my creative limits by thinking unique means to gather insights, to think creative and more importantly to "Think Human".

🧪

*Team Collaboration at it's peak 📈

We made a halloween presentation

Resume

Copyright © 2024 Shantanu Thorat

Made with love in Bloomington, USA