Reducing Cognitive Load with a Real-Time AI Sales Copilot
Imagine, as a salesperson, you had your own J.A.R.V.I.S. to help you across various sales pipeline stages
Product Name
ToMoBo
Company
Metaphy Labs (AI based B2B SaaS Startup)
My Role
Product Designer
Timeline
5 months
Status
Product development completed, project sunset due to market conditions
Scope
End-to-end product design including web dashboard, desktop app, landing page, email templates.
The Brief
Design an AI-powered sales copilot that provides real-time guidance during sales calls, helping sales executives qualify leads, answer objections, and capture insights all without disrupting the flow of conversation.
We explored various competitors and call recording tools such as fireflies.ai, meetgeek, otter.ai and found common pattern across them and listed their merits and demerits.
The Gap: No tool provides real-time, contextual guidance without disrupting the flow of conversation.
Worked closely with PM and founding team (which conducted initial sales executives interviews), we defined our north star:
Vision Statement: An AI copilot that sits alongside sales executives during calls, surfacing the right information at the right moment, without becoming a distraction.
Ambient intelligence
The AI should feel like a supportive team member, not an intrusive overlay. Suggestions appear contextually.
Reduce cognitive load, don't add to it
Every feature must simplify the rep's job. If it requires mental effort to use, it's failed.
Trust through transparency
Executives need to understand why the AI is making suggestions & show the reasoning
Designed for multitasking
Executives are on video calls. Interactions must be glanceable, scannable, and require minimal clicks.
This wasnt just about designing screens, it was about designing for an entirely new interaction paradigm.
Challenge 1: How to show AI suggestion without disrupting eye contact?
The Problem: Executives need to maintain eye contact and presence on video calls. If they are constantly looking down at a second screen, prospects notice
Challenge 2: What information is needed during vs after the call?
The Problem: Too much information during the call = overwhelm. Too little = missed opportunities
Challenge 3: How do you visualize sentiment analysis on audio recordings?
The Problem: Audio recordings can be 30 to 60 minutes long. How do you help reps quickly find the moments that matter forcing them to scrub through an entire transcript or listening to entire recording
When I first joined the project, the product vision was intentionally minimal, it just had a very basic flow which is below listed. This version was treated as the version 1 and in this sales executives just need to check email for the meeting analyis.
After validating the interviews we expanded our scope and redefined the userflows and included a dedicated Dashboard and Desktop app for sales executives with Hubspot CRM connection.
Onboarding and Connect Calendar
Meetings
Meetings - Upcoming meetings
Desktop App - ToMoBo Suggestions
Desktop App - Qualification Checklists
Meetings - Past Meetings
Homepage Dashboard
Settings & User Management
Email Templates
No product solves every problem perfectly. Here are the key trade-offs I navigated while designing Tomobo - what we prioritized, what we sacrificed and what I'd explore differently with more time
Trade-Off 1: Desktop-Only, No Mobile Companion App
We designed Tomobo exclusively for desktop (web dashboard + desktop app) with no mobile experience.
What we gained?
• Optimized interface for desktop
• Richer, more detailed analytics possible on larger screens
• Faster development timeline (single platform focus)
What we sacrificed?
• Sales reps who attended calls away from their desk can't access real-time suggestions
• No way to prep for meetings on-the-go
What I'd explored next:
A mobile companion app that provides:
• Lightweight prep view before meetings
• Audio-based suggestions via earbuds during calls
• Quick action items and follow ups on mobile
Trade-Off 2: Showing AI Confidence Levels
All AI suggestions appeared with equal visual weight & was not displaying any confidence indicators on AI suggestions
What we gained?
• Faster for executives to scan suggestions without evaluating confidence scores
• Cleaner, simple interface with less cognitive load
What we sacrificed?
• Risk of blindly following AI advice
• If AI hallucinates or misinterprets context, no warning signals
• Executives can't distinguish between high confidence vs uncertain suggestions
What I'd explored next:
Introduce confidence indicators with:
• Visual differentiation
• A/B test with real sales executives if confidence levels help in decision making
Based on user research and workflow analysis conducted by the Product Manager during discovery, we projected ToMoBo could improve sales process efficiency by approximately 65%. This projection was derived from:
Preparation efficiency
Reducing pre-call prep from 10 minutes to 2 minutes via automated context
Qualification improvement
Reducing missed questions from 60% to <10% through real time guidance
Time savings
Eliminating 15 - 20 minutes of manual CRM entry per call
If ToMoBo had shipped, we would have measured:
Actual time saved per call
Qualification completion rate
AI accuracy (percentage of suggestions marked helpful vs dismissed)
User adoption rate (percentage of scheduled calls where ToMoBo was active)
Around month 5 of development, a major competitor released a near-identical feature set with significant VC funding and enterprise sales motion. Our founding team made the difficult decision to pivot rather than compete in a saturated market.
But I'm proud of the work we accomplished in 6 months: a fully designed 0→1 product with complex, multi-surface interactions. The design thinking and problem-solving skills are what I carry forward.
This project taught me a lot of things not just about the product but also about startups and how things work at startups such as:
Market Timing matters as much as product quality
Startups are inherently risky
Speed is a competitive advantage
How to design for AI uncertainity
How to make strategic trade-offs under constraints
I'm presenting this case study not as "here's a product that succeeded," but as "here's how I think about complex problems, make design decisions, and learn from outcomes"
