Evergrowth
An AI-native GTM platform that enables companies to train intelligent AI agents using their own company knowledge, sales playbooks, customer personas, product documentation, and CRM data.
Unlike traditional CRMs, Evergrowth positions AI as a knowledgeable teammate that continuously learns and assists revenue teams throughout the sales lifecycle.
Role :
Founding UI/UX Designer
Duration:
8 Months
Industry :
Context
B2B sales teams deal with thousands of accounts and leads, but not every account is worth pursuing.
Sales reps need to research companies, identify the right contacts, validate ICP fit, find relevant buying signals, and decide who to reach out to all while keeping their CRM and sales processes up to date. Users could accomplish sophisticated tasks, but getting to the right feature often required understanding Evergrowth's internal terminology, navigating through several layers of the product, and understanding dependencies between different modules.
This created a fundamental product challenge:
Where I Came in :
As Evergrowth evolved into a more powerful GTM workspace, the growing number of features, workflows, and data points also made the product increasingly complex.
Rather than immediately redesigning the existing interface, I started by trying to understand how people actually experienced the platform. I conducted discovery interviews with internal users, power users, stakeholders, and existing clients, alongside a review of the existing product.
The goal was to identify recurring friction across the platform and understand which problems had the greatest impact on users.
Discovery Findings :
The platform lacked a clear mental model
Users frequently struggled to understand how different areas of Evergrowth were connected, particularly within the Agent Training Center. Feature placement, navigation, and terminology often made it unclear where to go next.
Users needed a faster path to value
Users wanted to spend less time figuring out the platform and more time getting their work done. Repetitive tasks and scattered entry points made onboarding and everyday workflows more time-consuming than necessary.
Powerful workflows came with unnecessary friction
Users relied heavily on agents, research, and workflows, but several processes required too many steps or an understanding of underlying system dependencies. There was also a need for better visibility and feedback around AI-generated outputs.
At the same time, stakeholders wanted users to have more visibility into how agents were performing and a way to provide feedback on their outputs.
This revealed two related opportunities: simplify complex workflows and create a continuous feedback loop for AI-generated outputs.
From Feedback to Action :
I consolidated the findings into a module-level UX audit, prioritizing each issue by severity and defining a proposed direction for improvement.
The priorities then shaped the redesign roadmap across information architecture, onboarding, core workflows, AI interactions, and the design system
Design Challenge 01 : Improving Overall Navigation
As the product matured, navigation expanded organically with the addition of new AI capabilities, automation tools, CRM modules, and reporting features. While existing users gradually adapted, first-time users often struggled to understand where information lived and how different modules related to one another.
Redesigning the Information Architecture
Problem
The existing navigation reflected the product's technical architecture rather than users' natural workflows. AI agents were scattered across different modules, forcing users to switch contexts depending on the task they wanted to complete.
Duplicated and similarly named elements, such as Industry Jargons, Persona Cards, and Personas, further created confusion about where information belonged. The Task Feed was also positioned within the sidebar despite primarily representing workflow activity.
Additionally, module-specific primary actions were placed within the global top navigation. This blurred the boundary between platform-wide navigation and actions relevant only to a specific module, making the hierarchy less predictable.
Together, these inconsistencies increased context switching and cognitive load, making the platform harder to navigate than necessary.
Design Approach
I restructured the information architecture around user goals instead of system architecture.
Modules were reorganized by these categories :
Intelligence Hub : Where AI learns, refines and Stores Intelligence.
Data Hub : Where the actual context is provided to these agents.
Automation Hub : Where users orchestrate how agents act on intelligence.
Related functionality was grouped together, reducing unnecessary navigation and improving wayfinding throughout the platform.
The revised architecture also established a clearer distinction between global navigation and contextual actions. Platform-wide elements remained in the global navigation, while primary actions specific to a module were moved into that module's own interface.
I also introduced a dedicated Agents module, bringing all 13 AI agents into one centralized space. Users could now discover and access the appropriate agent based on their goal without switching between different modules.
Outcome
The redesigned architecture gave users a clearer mental model of how the platform was structured and how different modules worked together.
By centralizing agents, separating global navigation from module-specific actions, and removing redundant structures, the experience became more workflow-oriented, predictable, and easier to navigate.
Design Challenge 02 : Reducing Onboarding Time for New Users
Problem
The existing dashboard functioned primarily as an information display rather than an operational workspace.Users frequently navigated through multiple screens to perform repetitive tasks, increasing interaction costs and slowing daily workflows.
Design Approach
The dashboard was redesigned to function as a personalized command center.
Several workflow improvements were introduced.
Saved Filters
Advanced filtering was frequently used by sales teams to segment prospects based on geography, engagement, industry, ownership, pipeline stage, and custom attributes including agents. Rather than requiring users to rebuild complex filter combinations each time, commonly used filter configurations could now be saved and surfaced directly on the dashboard.
With a single interaction, users could immediately access their target prospect lists.
Short-Form Video Tutorials
Because Evergrowth introduced a relatively complex ecosystem of AI agents and workflows, we introduced short-form, contextual video tutorials on the dashboard.
These provided users with quick guidance on key platform capabilities and helped them understand how to get started without relying entirely on customer-facing teams for onboarding and support.
Outcome
The redesigned dashboard significantly reduced the effort required to understand and start using the platform.
For one customer onboarding process, the support team previously required approximately 12–15 sessions to onboard a team. Following the redesign, this was reduced to approximately 6–9 sessions.
This meant users could understand the platform fundamentals and begin prospecting sooner, while customer-facing teams could spend less time on repetitive onboarding and more time supporting and working with additional clients.
Design Challenge 03 : Introducing In-App Feedback for Continuous AI Improvement
AI systems improve only when they receive high-quality feedback.
Although users regularly interacted with AI-generated recommendations, there was no structured way to communicate whether those responses were helpful or inaccurate.
Problem
Without an embedded feedback mechanism, valuable user insights remained disconnected from the AI improvement process.
This limited opportunities to refine recommendations and weakened user trust.
Design Approach
A lightweight in-app feedback system was introduced throughout AI-generated experiences. Users could quickly indicate whether responses were useful and provide contextual comments explaining why.
Outcome
The new feedback loop established an ongoing conversation between users and the AI system. Beyond improving future AI performance, it also increased user confidence by demonstrating that the platform continuously learns from real-world interactions.
Design Challenge 04 : Making Workflow Automation More Flexible
Workflow automation was one of Evergrowth’s most powerful capabilities, but the original builder was rigid and relied heavily on hidden upstream dependencies. Users often encountered automatically added agents without understanding why they were required, making workflows difficult to configure and modify.
Problem
The existing workflow builder forced users to think in terms of system dependencies rather than business logic. It lacked conditional logic, provided limited flexibility, and made it difficult to create or edit complex workflows.
Design Approach
Before redesigning the experience, I mapped the dependency structure between agents to understand how workflows were connected and identify opportunities to make these relationships more transparent.
The new builder introduced:
Drag-and-drop workflow creation for easier configuration.
Explicit dependency messaging to explain when additional agents were required and why.
Conditional logic to support more flexible workflows.
Agents and actions as modular building blocks.
Outcome
Automation became significantly more adaptable to changing business processes. Users could iterate on workflows more confidently while reducing the effort required to maintain increasingly sophisticated automation.
Impact
~50% Fewer Onboarding Sessions
Reduced the sessions required to onboard users by approximately half through a clearer information architecture and reorganized module structure.
60-80% Reduction in UX Support Tickets
Weekly UX-related tickets dropped from 5–6 to 1–2 following the product improvements.
2M+ Agent Runs every Month
After centralizing 13 agents into a dedicated Agent Workspace, the platform recorded 2M+ agent runs per month.
Beyond the redesign
Alongside restructuring the platform's information architecture and core workflows, I also contributed to several foundational product experiences that expanded the platform's capabilities.
CRM Integrations
I designed the CRM integration experience from the ground up, researching patterns from products such as Apollo and Outplay to understand how data moves between platforms. This helped establish a clearer framework for connecting external CRM data with Evergrowth's agent workflows.
Eva- The AI Workspace Assistant
I introduced Eva, an AI assistant designed to act as a central interface for managing and running agents within the workspace. Rather than requiring users to navigate between individual agents and modules, Eva provided a more conversational way to interact with the agent ecosystem.
