VIEW CASE STUDY
Rii LLM

What would you
like to know?

Case study

B2B SaaS

Agentic Ai X Salesforce

IRIS Visual Design

Period

3 months

Team

Ravikant Joshi

Pratik Dev

Riya Ghosh

Role

Product Designer

Skills

Brand Identity

Ai Flow Architecture

Product Strategy

Tools

Figma

Photoshop

Figma Make, Claude

Context

The average sales rep is buried in administrative work—updating CRMs, logging calls, finding data—which shatters their focus and kills sales momentum. The problem isn't just the loss of time; it's the loss of flow.

We didn't set out to build another productivity tool. We set out to design Rosa, an AI partner designed not just to manage data, but to anticipate the next best move, highlight the right opportunities, and clear the path for what sales reps do best: build connections and close deals

What if a sales rep could spend their entire day selling?

Solution

An AI agent that provides predictive guidance and automates the entire sales workflow, turning data into conversations.

I led the design of Rosa, a SaaS platform that acts as a strategic partner for sales representatives. Rosa integrates with a rep's core tools to deliver a prioritized daily brief, automate research, and assist in crafting personalized outreach. From surfacing the top 1% of leads to providing a "Deal Memory" for every interaction, Rosa is designed to be the engine for sales success.

The Problem

Sales reps are drowning in tasks, but starving for momentum.

My research into the sales workflow revealed a universal frustration. Reps like our persona, "Jordan," start their day facing a noisy pool of 500+ leads and a mountain of administrative tasks. This creates a reactive, inefficient, and demoralizing work environment.

Pain Points

Decision Fatigue

With hundreds of leads, reps don't know where to focus first. They waste the first, most productive hour of their day manually filtering lists instead of engaging high-intent prospects.

The "Research Tax"

Preparing for a single outreach call requires toggling across LinkedIn, Google News, industry blogs, and CRM history. This 1-hour, multi-tab research process for every key lead is a massive drain on productivity.

The Connection Gap

A fragmented workflow across multiple apps makes it impossible to have a single, unified view of a client. Key details from past interactions get lost, and outreach feels generic, not personal.

Key Insight

The solution wasn't a better dashboard. It was an intelligent agent that could perform the research and prioritization work for the rep, allowing them to start their day with action, not administration.

The Design Process

An AI-Augmented Workflow: From Concept to Prototype in Days, Not Weeks. To design a product centered on speed and intelligence, our process had to embody those same principles. Traditional design sprints felt too slow. Instead, I adopted a modern, AI-augmented workflow that leveraged generative AI as a foundational partner to accelerate the journey from a blank canvas to a testable prototype.

IRIS Outreach dashboard
IRIS video call UI

1. AI as the Starting Block

Instead of spending weeks on manual user flow diagrams and wireframes, I began by prompting AI with the core persona and problem. I used generative AI to: Brainstorm initial user flows for "a day in the life of a sales rep named Jordan." Generate multiple layout concepts for a proactive, AI-driven dashboard. Explore different information architectures for presenting complex data like lead scores and intent signals. This AI-first approach provided an incredible "base start," allowing us to bypass the most time-consuming parts of the initial ideation phase. It turned the "blank canvas problem" into a multiple-choice question, providing a rich set of concepts to refine.

2. Rapid Prototyping in Figma

Instead of spending weeks on manual user flow diagrams and wireframes, I began by prompting AI with the core persona and problem. I used generative AI to: Brainstorm initial user flows for "a day in the life of a sales rep named Jordan." Generate multiple layout concepts for a proactive, AI-driven dashboard. Explore different information architectures for presenting complex data like lead scores and intent signals. This AI-first approach provided an incredible "base start," allowing us to bypass the most time-consuming parts of the initial ideation phase. It turned the "blank canvas problem" into a multiple-choice question, providing a rich set of concepts to refine.

3. Designing the Core Agent: "Ask Rosa"

The centerpiece of the platform is "Ask Rosa," the central, conversational AI agent. This isn't a simple chatbot; it's designed to understand intent and execute multi-step tasks across integrated systems (CRM, email, calendar). The design process focused on mapping these agentic commands: Task Decomposition: we started by breaking down complex user requests into their fundamental actions. For example, the command "Get me prepped for my call with Delta Tech" was decomposed into a sequence: Query CRM for contact history → Scan Deal Memory for recent activity → Find related call recordings → Generate Smart Script with key talking points.

Start context
End context

4. Designing the "Sub-Agentic" Assistance

While "Ask Rosa" is the powerful command center, we also designed smaller, "sub-agentic" helpers that are embedded directly into the UI. These are proactive, contextual AI features that work silently in the background to automate specific tasks without needing an explicit command. The design involved identifying high-friction points in the workflow and embedding an AI solution: The "Lead Score" Agent constantly analyzes incoming leads, automatically scores them based on predefined criteria, and surfaces the justification for the score. The "Momentum" Agent monitors deal engagement. When it detects a key event (like a proposal being opened multiple times), it triggers the "Nudge" notification. The "CRM Hygiene" Agent scans for deals that haven't been updated and changes the priority status matrix to low priority.

Rosa sub-agentic assistance diagram

The Solution

A strategic partner that guides the entire sales motion. The Rosa platform is designed to be a rep's single source of truth and their primary strategic partner, turning hours of manual work into minutes of high-impact selling.

1. The Strategic Start

A strategic partner that guides the entire sales motion. The Rosa platform is designed to be a rep's single source of truth and their primary strategic partner, turning hours of manual work into minutes of high-impact selling. Jordan begins his day with Rosa's brief, instantly knowing which 7 of his 500 leads are showing the strongest intent signals, and which 3 stalled deals need immediate attention.

The Morning Brief
IRIS dashboard

2. The Intelligent Outreach

Jordan clicks on a high-score lead. Rosa provides the full research briefing and suggests a personalized email draft based on the lead's recent activity. Jordan tweaks it and sends. The entire flow, from qualification to outreach, takes under 10 minutes.

Intelligent Outreach — IRIS

3. The Momentum Nudge

Later, Rosa detects the lead has opened the proposal twice and sends a notification: "Proposal opened by Delta's VP twice in the past hour. Loop in AE before the window closes?" A deal acceleration moment is created.

Nudges
IRIS Assistive — Momentum

4. Assistive Canvas

Designing a modular, customizable dashboard that would serve as the rep's "mission control." We wanted to keep layout simple needs: Prioritized Tasks, At-Risk Deals, Key Metrics, and an AI Activity Feed. The key innovation was making these blocks both AI-curated and user-configurable, allowing reps to arrange their workspace to match their personal workflow.

IRIS Ask Rosa — Assistive Canvas

The Outcome

By automating admin work and providing predictive guidance, Rosa fundamentally changes the nature of the sales role.

Reflection

This project taught me that the future of enterprise AI isn't just automation, but anticipation. Designing Rosa required a shift from creating interfaces that display data to designing systems that provide predictive, actionable guidance. It's about empowering the user to make the best next move, every time.

The biggest design challenge was building trust in the AI's autonomy. A sales rep's relationships are their most valuable asset; they need to trust that the AI is enhancing, not risking, those relationships. We solved this by designing for a "glass box" AI: for every recommendation, Rosa shows the "why" behind it. For all external communication, we kept a "human-in-the-loop" model, where the rep gives the final approval.

If I were to continue this project, I would focus on the "AI Coaching" aspect. I'd explore using sentiment analysis on call recordings to provide reps with personalised feedback on their conversations, helping them improve their technique over time. This would evolve Rosa from a strategic partner into a true career coach, fully realising the vision of a system built for Momentum, Connection, and Success.