UX Design

Design Systems

Sales Industry

Designing a Scalable Enterprise Design System for AI Products

Designing a Scalable Enterprise Design System for AI Products

COmpany:

Proshort

Category:

UX Research and Design

My Role:

Lead Designer

Building a multi-product system that scaled across dashboards, AI workflows, CRM experiences, and multimodal interfaces — balancing consistency, flexibility, and future adaptability.

2025-2026

Timeline

AI-Ready

Impact

2.4× engagement

Impact

01 — THE REAL PROBLEM

Components Were Growing Faster Than the System

When the Platform Outpaces the Architecture

The platform evolved rapidly — from focused sales tooling into a broader AI-powered enterprise ecosystem. New workflows appeared constantly: AI coaching, multimodal feeds, CRM automation, analytics, roleplays, collaborative workflows. The issue wasn't lack of components. It was components evolving independently without a scalable system layer underneath.

Inconsistent spacing across 6 product surfaces. 14 button variants in production. 17 grey tokens with no semantic naming. Three modal patterns built by three different engineers in the same sprint. The more the platform scaled, the harder consistency became.

The Drift Audit

02 — WHAT MADE THIS HARD

This Wasn't Just a UI Library

Adaptive Behavioral Building Blocks

The challenge wasn't creating buttons and cards. The challenge was designing a system that could support structured and unstructured interfaces, adapt to AI-generated content, scale across enterprise workflows, maintain backward compatibility, and remain flexible without becoming inconsistent. Most components were no longer static UI elements. They became adaptive behavioral building blocks.

Backward compatibility while innovating

Challenge

AI-generated content with unpredictable length

New constraint

3 product squads, same component library

Scope

03 — SYSTEM THINKING

Designing Beyond Components

Five Layers Instead of One

Instead of designing isolated UI pieces, the system was reframed around five interconnected layers. Each layer serves the next — foundations give tokens their values, tokens give components their language, components combine into patterns, patterns compose into AI surfaces. Breaking any layer breaks everything above it.

FOUNDATIONS: Typography, spacing, color, elevation, motion → TOKENS: Primitive + semantic token architecture → COMPONENTS: Reusable scalable building blocks → PATTERNS: Enterprise workflows and interaction behaviors → AI SURFACES: Interfaces designed for dynamic generated content

System Architecture

04 — BUILDING FOR SCALE

Two Realities at Once

Stability vs Scalability vs Adaptability

Existing enterprise products already depended on older tokens and legacy behaviors. New AI-native workflows required flexibility that the old system never anticipated. This created constant tension between stability, scalability, and adaptability. The goal became: evolve the system without breaking trust.

Evolve the system without breaking trust. Every change had to be invisible to the end user and painless for the engineer.

Depended on older tokens — cannot break

Current products

Depended on older tokens — cannot break

Current products

Required flexibility old system couldn't anticipate

Future products

Required flexibility old system couldn't anticipate

Future products

05 — COMPONENT ECOSYSTEM

A Unified Language Across Products

From Atoms to Enterprise Patterns

The system organised components into three tiers — not by visual complexity, but by how much context they required to use correctly. Foundations require no context. Core components require product context. Advanced patterns require workflow context. The tier determined documentation depth and governance requirements.

Typography · Colors · Spacing · Elevation · Grid

Foundations

Typography · Colors · Spacing · Elevation · Grid

Foundations

Tables · Tags · Inputs · Buttons · Nav · Filters · Modals

Core Components

Tables · Tags · Inputs · Buttons · Nav · Filters · Modals

Core Components

AI feeds · Coaching systems · CRM intelligence · Roleplay workflows

Advanced Patterns

AI feeds · Coaching systems · CRM intelligence · Roleplay workflows

Advanced Patterns

06 — AI CHANGED THE RULES

Human-Written UI Became Machine-Generated UI

From Visual Consistency to Resilient Interface Behavior

Traditional design systems assume predictable content. AI interfaces broke those assumptions. The system now had to support unpredictable text lengths, generated labels, dynamic hierarchy, contextual metadata, and adaptive layouts. This forced the design system to evolve from visual consistency to resilient interface behavior.

Static component → Behavioral container. Fixed sizing → Adaptive sizing. Controlled content → AI-generated content. Single-context usage → Multi-context semantic layer. The tag component became the clearest example of this evolution.

What AI Changed

07 — DEEP DIVE

One Component Explained the Entire System

The Tag That Became a Semantic Enterprise Layer

Among all components, one became the clearest example of system evolution: the Tag component. What started as a simple dashboard label evolved into a semantic enterprise communication layer used across dashboards, AI feeds, CRM workflows, skill systems, activity timelines, company identifiers, and follow-up systems. Four versions. One architectural lesson.

We thought we were designing a label. We were actually designing a semantic communication protocol for the entire product.

Static dashboard labels

V1

Icon-based semantic tags

V2

Responsive enterprise tags

V3

AI-compatible semantic containers

V4

Tag System Case Study →

How a dashboard label became an enterprise semantic system

08 — KEY LEARNINGS

What This Taught Me

Three Principles That Survived the Build

01

Consistency alone doesn't scale systems

Flexibility without governance creates fragmentation. The two have to be designed together from the first token decision.

02

AI changes component assumptions

Generated content introduces behavioral unpredictability traditional systems aren't designed for. Every component needs an AI content test before it ships.

03

Design systems are organizational systems

The hardest problems weren't visual — they were coordination, scalability, and long-term adaptability. The system succeeded when engineers adopted it faster than they built around it.

Let's Talk

Let's Talk

Let's Talk

Let's Talk

Scribble Arrow

(book a call)

UX Design

Design Systems

Sales Industry

Designing a Scalable Enterprise Design System for AI Products

Designing a Scalable Enterprise Design System for AI Products

COmpany:

Proshort

Category:

UX Research and Design

My Role:

Lead Designer

Building a multi-product system that scaled across dashboards, AI workflows, CRM experiences, and multimodal interfaces — balancing consistency, flexibility, and future adaptability.

2025-2026

Timeline

AI-Ready

Impact

2.4× engagement

Impact

01 — THE REAL PROBLEM

Components Were Growing Faster Than the System

When the Platform Outpaces the Architecture

The platform evolved rapidly — from focused sales tooling into a broader AI-powered enterprise ecosystem. New workflows appeared constantly: AI coaching, multimodal feeds, CRM automation, analytics, roleplays, collaborative workflows. The issue wasn't lack of components. It was components evolving independently without a scalable system layer underneath.

Inconsistent spacing across 6 product surfaces. 14 button variants in production. 17 grey tokens with no semantic naming. Three modal patterns built by three different engineers in the same sprint. The more the platform scaled, the harder consistency became.

The Drift Audit

02 — WHAT MADE THIS HARD

This Wasn't Just a UI Library

Adaptive Behavioral Building Blocks

The challenge wasn't creating buttons and cards. The challenge was designing a system that could support structured and unstructured interfaces, adapt to AI-generated content, scale across enterprise workflows, maintain backward compatibility, and remain flexible without becoming inconsistent. Most components were no longer static UI elements. They became adaptive behavioral building blocks.

Backward compatibility while innovating

Challenge

AI-generated content with unpredictable length

New constraint

3 product squads, same component library

Scope

03 — SYSTEM THINKING

Designing Beyond Components

Five Layers Instead of One

Instead of designing isolated UI pieces, the system was reframed around five interconnected layers. Each layer serves the next — foundations give tokens their values, tokens give components their language, components combine into patterns, patterns compose into AI surfaces. Breaking any layer breaks everything above it.

FOUNDATIONS: Typography, spacing, color, elevation, motion → TOKENS: Primitive + semantic token architecture → COMPONENTS: Reusable scalable building blocks → PATTERNS: Enterprise workflows and interaction behaviors → AI SURFACES: Interfaces designed for dynamic generated content

System Architecture

04 — BUILDING FOR SCALE

Two Realities at Once

Stability vs Scalability vs Adaptability

Existing enterprise products already depended on older tokens and legacy behaviors. New AI-native workflows required flexibility that the old system never anticipated. This created constant tension between stability, scalability, and adaptability. The goal became: evolve the system without breaking trust.

Evolve the system without breaking trust. Every change had to be invisible to the end user and painless for the engineer.

Depended on older tokens — cannot break

Current products

Depended on older tokens — cannot break

Current products

Required flexibility old system couldn't anticipate

Future products

Required flexibility old system couldn't anticipate

Future products

05 — COMPONENT ECOSYSTEM

A Unified Language Across Products

From Atoms to Enterprise Patterns

The system organised components into three tiers — not by visual complexity, but by how much context they required to use correctly. Foundations require no context. Core components require product context. Advanced patterns require workflow context. The tier determined documentation depth and governance requirements.

Typography · Colors · Spacing · Elevation · Grid

Foundations

Typography · Colors · Spacing · Elevation · Grid

Foundations

Tables · Tags · Inputs · Buttons · Nav · Filters · Modals

Core Components

Tables · Tags · Inputs · Buttons · Nav · Filters · Modals

Core Components

AI feeds · Coaching systems · CRM intelligence · Roleplay workflows

Advanced Patterns

AI feeds · Coaching systems · CRM intelligence · Roleplay workflows

Advanced Patterns

06 — AI CHANGED THE RULES

Human-Written UI Became Machine-Generated UI

From Visual Consistency to Resilient Interface Behavior

Traditional design systems assume predictable content. AI interfaces broke those assumptions. The system now had to support unpredictable text lengths, generated labels, dynamic hierarchy, contextual metadata, and adaptive layouts. This forced the design system to evolve from visual consistency to resilient interface behavior.

Static component → Behavioral container. Fixed sizing → Adaptive sizing. Controlled content → AI-generated content. Single-context usage → Multi-context semantic layer. The tag component became the clearest example of this evolution.

What AI Changed

07 — DEEP DIVE

One Component Explained the Entire System

The Tag That Became a Semantic Enterprise Layer

Among all components, one became the clearest example of system evolution: the Tag component. What started as a simple dashboard label evolved into a semantic enterprise communication layer used across dashboards, AI feeds, CRM workflows, skill systems, activity timelines, company identifiers, and follow-up systems. Four versions. One architectural lesson.

We thought we were designing a label. We were actually designing a semantic communication protocol for the entire product.

Static dashboard labels

V1

Icon-based semantic tags

V2

Responsive enterprise tags

V3

AI-compatible semantic containers

V4

Tag System Case Study →

How a dashboard label became an enterprise semantic system

08 — KEY LEARNINGS

What This Taught Me

Three Principles That Survived the Build

01

Consistency alone doesn't scale systems

Flexibility without governance creates fragmentation. The two have to be designed together from the first token decision.

02

AI changes component assumptions

Generated content introduces behavioral unpredictability traditional systems aren't designed for. Every component needs an AI content test before it ships.

03

Design systems are organizational systems

The hardest problems weren't visual — they were coordination, scalability, and long-term adaptability. The system succeeded when engineers adopted it faster than they built around it.

Let's Talk

Let's Talk

Let's Talk

Let's Talk

Scribble Arrow

(book a call)

UX Design

Design Systems

Sales Industry

Designing a Scalable Enterprise Design System for AI Products

Designing a Scalable Enterprise Design System for AI Products

COmpany:

Proshort

Category:

UX Research and Design

My Role:

Lead Designer

Building a multi-product system that scaled across dashboards, AI workflows, CRM experiences, and multimodal interfaces — balancing consistency, flexibility, and future adaptability.

2025-2026

Timeline

AI-Ready

Impact

2.4× engagement

Impact

01 — THE REAL PROBLEM

Components Were Growing Faster Than the System

When the Platform Outpaces the Architecture

The platform evolved rapidly — from focused sales tooling into a broader AI-powered enterprise ecosystem. New workflows appeared constantly: AI coaching, multimodal feeds, CRM automation, analytics, roleplays, collaborative workflows. The issue wasn't lack of components. It was components evolving independently without a scalable system layer underneath.

Inconsistent spacing across 6 product surfaces. 14 button variants in production. 17 grey tokens with no semantic naming. Three modal patterns built by three different engineers in the same sprint. The more the platform scaled, the harder consistency became.

The Drift Audit

02 — WHAT MADE THIS HARD

This Wasn't Just a UI Library

Adaptive Behavioral Building Blocks

The challenge wasn't creating buttons and cards. The challenge was designing a system that could support structured and unstructured interfaces, adapt to AI-generated content, scale across enterprise workflows, maintain backward compatibility, and remain flexible without becoming inconsistent. Most components were no longer static UI elements. They became adaptive behavioral building blocks.

Backward compatibility while innovating

Challenge

AI-generated content with unpredictable length

New constraint

3 product squads, same component library

Scope

03 — SYSTEM THINKING

Designing Beyond Components

Five Layers Instead of One

Instead of designing isolated UI pieces, the system was reframed around five interconnected layers. Each layer serves the next — foundations give tokens their values, tokens give components their language, components combine into patterns, patterns compose into AI surfaces. Breaking any layer breaks everything above it.

FOUNDATIONS: Typography, spacing, color, elevation, motion → TOKENS: Primitive + semantic token architecture → COMPONENTS: Reusable scalable building blocks → PATTERNS: Enterprise workflows and interaction behaviors → AI SURFACES: Interfaces designed for dynamic generated content

System Architecture

04 — BUILDING FOR SCALE

Two Realities at Once

Stability vs Scalability vs Adaptability

Existing enterprise products already depended on older tokens and legacy behaviors. New AI-native workflows required flexibility that the old system never anticipated. This created constant tension between stability, scalability, and adaptability. The goal became: evolve the system without breaking trust.

Evolve the system without breaking trust. Every change had to be invisible to the end user and painless for the engineer.

Depended on older tokens — cannot break

Current products

Depended on older tokens — cannot break

Current products

Required flexibility old system couldn't anticipate

Future products

Required flexibility old system couldn't anticipate

Future products

05 — COMPONENT ECOSYSTEM

A Unified Language Across Products

From Atoms to Enterprise Patterns

The system organised components into three tiers — not by visual complexity, but by how much context they required to use correctly. Foundations require no context. Core components require product context. Advanced patterns require workflow context. The tier determined documentation depth and governance requirements.

Typography · Colors · Spacing · Elevation · Grid

Foundations

Typography · Colors · Spacing · Elevation · Grid

Foundations

Tables · Tags · Inputs · Buttons · Nav · Filters · Modals

Core Components

Tables · Tags · Inputs · Buttons · Nav · Filters · Modals

Core Components

AI feeds · Coaching systems · CRM intelligence · Roleplay workflows

Advanced Patterns

AI feeds · Coaching systems · CRM intelligence · Roleplay workflows

Advanced Patterns

06 — AI CHANGED THE RULES

Human-Written UI Became Machine-Generated UI

From Visual Consistency to Resilient Interface Behavior

Traditional design systems assume predictable content. AI interfaces broke those assumptions. The system now had to support unpredictable text lengths, generated labels, dynamic hierarchy, contextual metadata, and adaptive layouts. This forced the design system to evolve from visual consistency to resilient interface behavior.

Static component → Behavioral container. Fixed sizing → Adaptive sizing. Controlled content → AI-generated content. Single-context usage → Multi-context semantic layer. The tag component became the clearest example of this evolution.

What AI Changed

07 — DEEP DIVE

One Component Explained the Entire System

The Tag That Became a Semantic Enterprise Layer

Among all components, one became the clearest example of system evolution: the Tag component. What started as a simple dashboard label evolved into a semantic enterprise communication layer used across dashboards, AI feeds, CRM workflows, skill systems, activity timelines, company identifiers, and follow-up systems. Four versions. One architectural lesson.

We thought we were designing a label. We were actually designing a semantic communication protocol for the entire product.

Static dashboard labels

V1

Icon-based semantic tags

V2

Responsive enterprise tags

V3

AI-compatible semantic containers

V4

Tag System Case Study →

How a dashboard label became an enterprise semantic system

08 — KEY LEARNINGS

What This Taught Me

Three Principles That Survived the Build

01

Consistency alone doesn't scale systems

Flexibility without governance creates fragmentation. The two have to be designed together from the first token decision.

02

AI changes component assumptions

Generated content introduces behavioral unpredictability traditional systems aren't designed for. Every component needs an AI content test before it ships.

03

Design systems are organizational systems

The hardest problems weren't visual — they were coordination, scalability, and long-term adaptability. The system succeeded when engineers adopted it faster than they built around it.

Let's Talk

Let's Talk

Let's Talk

Let's Talk

Scribble Arrow

(book a call)