Project Overview
As Senior Product Researcher at Vivun, I led research and design efforts to reimagine how Sales Engineers log their activities. Vivun's platform helps Sales Engineers track deal pipelines and uses AI-powered "Hero Scores" to predict deal success. However, the accuracy of these scores depended on consistent activity logging—a task users found burdensome and often skipped.
Problem Statement
To provide an accurate scoring system for deals in a pipeline, we need to collect metadata from the activities of a Sales Engineer. However, low activity logging rates were undermining the Hero Score's accuracy, creating a negative feedback loop where inaccurate scores further demotivated logging.
Context & Challenge
Business Context
Vivun's Hero Score was a key differentiator—an AI-powered tool that analyzed Sales Engineer activities to predict which deals were most likely to close. But the AI was only as good as the data it received. Sales Engineers were logging activities inconsistently, if at all, leading to:
- Inaccurate deal scoring and predictions
- Reduced trust in the Hero Score feature
- Missed opportunities for coaching and improvement
- Incomplete deal pipeline visibility for managers
User Pain Points
"I'm already spending my day in back-to-back customer calls. The last thing I want to do is spend another 30 minutes logging what I just did. It feels like unpaid administrative work." — Senior Sales Engineer, Enterprise Software Company
Key challenges identified:
- Cognitive burden: Users had to recall and describe activities after busy days
- Time constraints: Activity logging felt like non-revenue-generating work
- Unclear value: Users didn't see immediate benefits from logging
- Repetitive data entry: Similar activities required re-typing the same information
Constraints
- Users wouldn't tolerate additional time spent on logging
- Solution needed to work across mobile and desktop
- Required integration with existing Hero Score AI system
- Had to maintain data quality standards for accurate scoring
Research Process
Research Approach
I conducted continuous user research over 10 weeks, combining multiple methods to deeply understand user workflows and pain points:
Weekly Customer Conversations (n=20+)
Held regular conversations with Sales Engineers to understand their daily workflows, logging habits, and frustrations. These sessions revealed when and how logging created friction in their day.
Workflow Shadowing (n=5)
Observed Sales Engineers throughout their workday to understand when activities occurred and what context was available for logging. Identified key moments where logging could be seamlessly integrated.
Usage Analytics Review
Analyzed platform data to identify logging patterns, drop-off points, and correlations between logging frequency and Hero Score accuracy.
Key Research Questions
- At what point in their workflow do Sales Engineers currently log activities?
- What information is hardest to recall or articulate when logging?
- What motivates users to log activities (when they do)?
- How could we reduce the cognitive load of activity description?
- What would make logging feel valuable rather than burdensome?
Key Research Insights
1. Logging Happens in Brief Moments Between Calls
Sales Engineers don't have dedicated time for logging. They log activities in 30-second to 2-minute windows between meetings, often on mobile devices while walking to their next meeting or waiting for calls to start.
"I have maybe a minute or two between calls. If logging takes longer than that, I just skip it and tell myself I'll do it later—but later never comes." — Sales Engineer, Tech Startup
2. Describing Activities is the Biggest Bottleneck
Users spent most of their logging time trying to articulate what happened in a call or meeting. The blank text field was intimidating and required significant mental effort to compose clear descriptions.
- Average time spent writing activity descriptions: 90 seconds
- Most users struggled with how detailed to be
- Many activities followed predictable patterns but still required custom descriptions
3. Sentiment and Tone Were Hard to Capture
The Hero Score AI needed to understand not just what activities occurred, but the sentiment around them (positive customer response, pushback, enthusiasm, etc.). Users found it difficult to quantify qualitative impressions.
"I know the customer was really excited about the demo, but how do I put that into the logging form? There's no field for 'customer vibes.'" — Senior Sales Engineer, Enterprise SaaS
4. Users Craved Speed and Predictability
When asked what would make logging better, users consistently requested:
- Pre-written phrases they could select instead of typing
- AI suggestions based on calendar events and past patterns
- Quick shortcuts for common activity types
- Voice input for hands-free logging
Design Solutions
Solution 1: AI-Generated Activity Descriptions
Leveraged AI to generate suggested activity descriptions based on calendar metadata, deal context, and historical patterns. Users could accept, edit, or replace suggestions with a single tap.
AI-generated suggestions dramatically reduced typing and mental effort
How It Works
- System detects calendar event and deal context
- AI generates 2-3 relevant activity descriptions
- User selects best match or edits for accuracy
- System learns from user selections over time
Solution 2: Automated Sentiment Analysis
Introduced sentiment indicators that users could quickly select (Positive, Neutral, Negative, Mixed) instead of trying to describe emotional tone in text. The Hero Score AI used these signals to better predict deal health.
One-tap sentiment selection replaced lengthy text descriptions
Design Rationale
Rather than asking users to write "The customer seemed enthusiastic," we let them select a sentiment indicator. This was faster for users and more consistent for the AI to process.
Solution 3: Pre-Packaged Phrase Library
Created a library of common phrases and activity patterns that users could quickly select and combine, like building blocks. Reduced typing by up to 75% for standard activities.
Common Phrases
- "Demonstrated [feature] capabilities"
- "Addressed technical questions about [topic]"
- "Walked through integration requirements"
- "Discussed security and compliance"
Outcome Indicators
- "Customer ready to move forward"
- "Waiting on customer decision"
- "Requested follow-up demo"
- "Identified blockers"
Solution 4: Voice-Activated Logging (Hackathon Prototype)
During a company hackathon, I prototyped "Ava," a voice-activated AI assistant for activity logging. Sales Engineers could verbally describe their activities while walking between meetings, and Ava would structure the data for the platform.
Ava voice assistant prototype for hands-free activity logging
"Being able to just say 'Hey Ava, log my demo with Acme Corp—customer loved the integration features' while I'm walking to my next meeting is a game-changer." — Beta Tester, Sales Engineer
Design Process
Discovery & Research
Conducted user interviews, workflow shadowing, and analyzed usage data to identify pain points
Ideation & Wireframing
Sketched concepts in Miro, explored AI-powered solutions, created low-fidelity prototypes
High-Fidelity Design & Testing
Built interactive prototypes in Figma, conducted usability tests with 8 users, iterated based on feedback
Implementation & Validation
Worked with engineering to implement designs, validated with beta users, measured impact
Impact & Results
Quantitative Results
Qualitative Feedback
"The new logging experience is night and day. I actually use it now because it takes seconds instead of minutes. The AI suggestions are scarily accurate—it's like it knows what I'm going to say." — Sales Engineer, Fortune 500 Company
"My team's Hero Scores are finally reliable now that everyone is logging consistently. I can actually use them to coach and prioritize deals." — VP of Sales Engineering
Business Impact
- Hero Score accuracy improved by 34% due to more consistent and comprehensive activity data
- User satisfaction with logging increased from 2.1/5 to 4.3/5
- Competitive advantage: Activity logging became a product differentiator vs. competitors
- Reduced churn: Sales Engineers cited improved logging experience as reason to stay with platform
Before/After Comparison
Activity logging experience: manual entry (before) vs. AI-assisted (after)
Reflection & Learnings
What Went Well
- Continuous user engagement: Weekly conversations with Sales Engineers kept research grounded in real workflows and pain points
- AI as enabler: Leveraging AI to reduce friction rather than add complexity was the key to adoption
- Rapid prototyping: The Ava voice assistant hackathon prototype validated future directions and excited users
- Measurable impact: Clear metrics (logging rate, time spent, accuracy) allowed us to demonstrate value
What I'd Do Differently
- Earlier cross-platform testing: Should have tested mobile experience earlier since most logging happens on phones
- Broader sentiment categories: Users wanted more nuanced sentiment options beyond positive/neutral/negative
- Manager perspective: Could have spent more time understanding how managers use activity data for coaching
Skills Developed
- Deepened expertise in designing AI-powered user experiences
- Learned to balance AI automation with user control and transparency
- Improved ability to design for micro-moments and mobile-first workflows
- Strengthened skills in voice interface design and conversational UX
Key Takeaways
On AI in UX: AI is most powerful when it reduces friction rather than adding features. Users don't want AI for AI's sake—they want tasks to be faster and easier. The AI-generated descriptions succeeded because they eliminated work, not because users were excited about AI.
On user motivation: Users will adopt features that save them time, even if those features require trusting AI. But that trust must be earned through consistent accuracy and the ability to override suggestions.
On voice interfaces: The Ava prototype showed that voice is ideal for hands-free, mobile-first scenarios. But voice must be optional—some users prefer visual interfaces even when mobile.