Industry Sponsored
Conversational AI
AI Chatbot
Optimizing the Salesforce Chatbot for Faster Navigation and Higher Engagement
My Role
Contextual Design
User Research
Interaction Design
Prototyping
Usability Test
Project
Industry sponsored project by Salesforce
Timeline
5 months
Aug 2024 - Dec 2024
Research Methods
Competitive Analysis
Contextual Inquiries
Affinity Mapping
Design Critiques with Internal Stakeholders
Usability Testing
Overview
Context
When chatbots first became popular in e-commerce and SaaS, they promised a seamless, human-like experience that would help users navigate complex systems with ease. However, Salesforce’s chatbot, designed to help businesses discover products, fell short of expectations and introduced unnecessary friction.
Project Goal
Designing a human-centered AI chatbot experience for businesses using the Salesforce website with a specific focus on product discovery.
Research
Key Research Methods
My research process was rooted in understanding user behavior and identifying key pain points through direct observation, competitor analysis, and iterative testing.
🔍
Competitive Analysis & Contextual Inquiries
Researched Amazon Rufus, ChatGPT, and Einstein Assistant to identify gaps in conversational AI capabilities.
🗣
User Testing
Conducted multiple rounds of usability testing to uncover user frustrations and validate design decisions.
📌
Industry Feedback & Iteration
Engaged in 15+ collaborative sessions with mentors and Salesforce professionals, refining the design through multiple critique rounds.
Problem
Why Users Struggled
Our research revealed key pain points that hindered the chatbot’s effectiveness:
Key Issues
Design Solution
A Smarter, Context-Aware Chatbot
To address these challenges, we introduced a structured set of features designed to create a seamless, intuitive, and engaging experience. These improvements fall into four key areas:
01.
Establishing Personality & Tone
We created a chatbot persona that aligns with Salesforce’s brand, making interactions feel more engaging and natural.
⛔
Problem
The chatbot felt robotic and disengaging, making interactions impersonal.
✅
Solution
Designed a professional yet approachable chatbot persona aligned with Salesforce’s brand.
👉
Impact
More engaging interactions, reducing user frustration and increasing usability.
02.
Enhancing Conversations & Assistance
We refined the chatbot’s interactions to make them feel more intuitive and engaging.
⛔
Problem
Users struggled to phrase queries, leading to vague or irrelevant responses.
✅
Solution
Introduced clarifying questions and suggestive prompts to guide users.
👉
Impact
More relevant responses, improved engagement, and reduced drop-offs.
📝 Suggestive Prompt
When users hesitate or delete text, the chatbot suggests ways to phrase their questions more effectively.
💡 Clarifying Questions
Instead of responding vaguely, the chatbot asks targeted follow-ups to refine user queries
💤 Handling User Inactivity
Detects inactivity and nudges users with reminders or relevant suggestions to keep them engaged.
03.
Streamlining Product Discovery & Decision-Making
To bridge the gap between chatbot interaction and browsing experience: Users needed a structured way to browse products without endless scrolling.
⛔
Problem
Users found it difficult to compare products and lost track of key details.
✅
Solution
Implemented structured product comparisons and a chatbot timeline for easy reference.
👉
Impact
Faster decision-making, reduced cognitive load, and minimized scrolling.
📊 Comparing Products
The chatbot organizes product details into structured comparison tables, eliminating the need to search through long chat histories.
📈 Chat Timeline
A persistent chat progress bar enables users to jump between topics, track decisions, and revisit past recommendations.
04.
Enabling Smooth Transitions & Support
To ensure users receive help at the right time without disruption
⛔
Problem
Users feared losing chat history while browsing and received agent support too soon.
✅
Solution
Integrated chatbot redirection and improved agent handoff mechanisms.
👉
Impact
Seamless chat access across pages and smarter, context-aware agent support.
Chatbot Redirection
Allows users to switch between chatbot and website without losing their conversation history.
Connect to an Agent
Users can request human support at the right time instead of being forced into premature interactions.
Impact
Increased Engagement, Faster Decisions, and Smarter Support
Our refinements made the chatbot easier to navigate, more engaging, and user-driven. Users now had better control, clearer guidance, and a chatbot that adapted to their needs
📉
Decrease in user drop-off
More users completed their product exploration without frustration.
📈
Higher engagement
Users interacted more with the chatbot due to improved guidance and structured conversations.
⚡
Faster decision-making
Reduced scrolling time, leading to quicker product evaluations.
👤
Better sales handoff
Users were more prepared before connecting with a sales rep, improving conversion rates.
Iterations
Understanding Users
To create a truly human-centered AI, I needed to dig deep into how users interact with chatbots and where traditional experiences fall short. My research was guided by three main questions:
How do users expect chatbots to behave?
What makes chatbot interactions frustrating or useful?
How can we design an AI assistant that feels more intuitive?
Research Approaches
AI Simulation for Deeper Insights
To understand Einstein Assistant’s limitations, we took an innovative approach: using ChatGPT to simulate its responses.
This allowed us to observe real user frustrations in a controlled environment.
Behavioral Patterns Identified
Through testing, we uncovered three major user struggles:
1.
Hesitation to start conversations
Users were unsure how to phrase their queries.
2.
Unstructured inputs
Users provided vague or incomplete questions, leading to irrelevant responses.
3.
Overwhelming responses
Long chatbot replies frustrated users, causing them to abandon interactions.
Transforming Insights into Actionable Solutions
By focusing on real user behaviors, I transformed broad frustrations into targeted solutions:
Here’s how we addressed major challenges:
Long, unstructured responses
Timeline feature
Hesitation in phrasing queries
Suggestive Prompts
Robotic & impersonal chatbot
Improvements in Personality & Tone
Reflection
What I Learned
This project reinforced the importance of context-awareness in AI-driven experiences. A chatbot isn’t just a tool, it’s a guide that should reduce user effort and create meaningful interactions.
1.
Iteration is everything!
Each feature changed drastically after user feedback.
2.
AI should guide, NOT assume
Users don’t always know how to ask questions.
3.
Small UX tweaks = big wins
The timeline feature alone doubled usability.
Next Steps
Looking forward, I want to:
🗣
Expand multimodal interactions
Integrate voice, visuals, and text AI capabilities
🧠
Media Integration
Adding videos inside chat for richer product demos.
📊
Returning User Experience
Customizing responses for returning users.
Final Steps
This project taught me that designing a chatbot isn’t just about improving responses, it’s about crafting an experience that feels seamless, human, and genuinely useful. By refining tone, guiding conversations, and reducing friction, I transformed a frustrating tool into an intuitive assistant.
Conversational AI has so much potential to grow. Whether it's improving memory retention, integrating multimodal interactions, or making responses even more adaptive, there’s always more to explore. I’m excited to keep pushing the boundaries of what AI can do to enhance user experiences.
At its core, the best AI doesn’t just provide answers. It understands, anticipates, and supports users every step of the way. That’s the kind of impact I strive to create in every project.