LESSEN LLC

2026

Developed scalable AI design system for multi-product ecosystem

Responsibilities

Design System · UX Strategy · System Architecture · AI Interaction Design · Research Synthesis

Project Scope

B2B · B2C · B2B2C · AI Platform · Desktop & Mobile · AI Interaction Layer on Existing Design System

Team

Solo Designer · PM

Timeline

2 months

The Brief

AI was introduced piecemeal across chat and workflows without a shared interaction model, leading to fragmented patterns and unclear boundaries of use. This project defined a unified AI interaction system and architecture layer to standardize how AI is embedded and behaves across the product.

Intent

The goal was to design an AI interaction architecture grounded in real usage patterns rather than default chat‑first assumptions. The system needed to standardize how AI appears across workflows and chat, and give teams a shared framework to scope, compare, and evaluate new AI features.

IMPACT

Established a unified AI interaction system across chat and embedded workflows, creating a consistent model for how AI is surfaced and integrated across the product ecosystem. This reduced ambiguity for teams designing new AI features and introduced a shared structure for evaluating, prioritizing, and building AI experiences.

System Design

New HCAI Framework

System Components

30+ Components

System Scope

6 B2B Products

APPROACH

01

Audit

Audit system reality – Mapping AI usage and fragmentation

Reviewed existing AI usage across chat and embedded surfaces. Most interactions were short, typically 1–2 exchanges, with users quickly returning to their primary workflow. AI entry points were inconsistent, creating friction in discovery and repeat usage.

01

Audit

Audit system reality – Mapping AI usage and fragmentation

Reviewed existing AI usage across chat and embedded surfaces. Most interactions were short, typically 1–2 exchanges, with users quickly returning to their primary workflow. AI entry points were inconsistent, creating friction in discovery and repeat usage.

02

Structure

Structure system foundations – Defining an embedded AI interaction layer

Translated usage constraints into a dedicated AI interaction layer within the existing design system. This shifted focus away from chat‑first experiences toward lightweight, contextual AI embedded directly within workflows, with shared rules for placement, entry points, and reusable UI patterns.

02

Structure

Structure system foundations – Defining an embedded AI interaction layer

Translated usage constraints into a dedicated AI interaction layer within the existing design system. This shifted focus away from chat‑first experiences toward lightweight, contextual AI embedded directly within workflows, with shared rules for placement, entry points, and reusable UI patterns.

03

Behavior Design

Defining interaction patterns and the AICM

Mapped high‑value AI use cases, particularly summarization, work order creation, and contextual insights. From this, two interaction types were defined: Workflow AI – bounded, task‑driven interactions embedded in product flows Agent AI – open‑ended, conversational experiences

03

Behavior Design

Defining interaction patterns and the AICM

Mapped high‑value AI use cases, particularly summarization, work order creation, and contextual insights. From this, two interaction types were defined: Workflow AI – bounded, task‑driven interactions embedded in product flows Agent AI – open‑ended, conversational experiences

04

Operationalize System

Documenting patterns and aligning teams

Documented the AI interaction layer as part of the design system, including pattern guidelines, decision criteria for Workflow vs Agent AI, and example flows. Aligned PM and engineering on when to use each pattern, reducing one‑off AI feature requests and inconsistencies.

04

Operationalize System

Documenting patterns and aligning teams

Documented the AI interaction layer as part of the design system, including pattern guidelines, decision criteria for Workflow vs Agent AI, and example flows. Aligned PM and engineering on when to use each pattern, reducing one‑off AI feature requests and inconsistencies.

04

Scale System

Building reusable AI components

Delivered ~30 reusable components and variants across chat, embedded insights, and workflow actions. Components were designed for reuse across AI use cases while maintaining consistency in how AI value is surfaced and embedded across the product.

05

Scale System

Building reusable AI components

Delivered ~30 reusable components and variants across chat, embedded insights, and workflow actions. Components were designed for reuse across AI use cases while maintaining consistency in how AI value is surfaced and embedded across the product.

SOLUTION

A unified AI interaction system that defines how AI is embedded into workflows and how it behaves across chat. The system separates reusable UI patterns from interaction types, allowing teams to design consistent AI experiences without redefining behavior for each new feature.

01

Embedded-first AI components

Shifted AI from a chat‑centric interface to contextual, in‑flow support within workflows.

01

Embedded-first AI components

Shifted AI from a chat‑centric interface to contextual, in‑flow support within workflows.

02

Componentized AI interaction layer

Created reusable components focused on high‑value AI tasks like summarization, creation, and contextual insights, enabling consistent application across products

02

Componentized AI interaction layer

Created reusable components focused on high‑value AI tasks like summarization, creation, and contextual insights, enabling consistent application across products

03

HCAI Framework

Introduced a classification framework separating Workflow AI (bounded, task‑based interactions) from Agent AI (open‑ended conversational interactions), providing a shared decision model for how AI should be designed and embedded.

RETROSPECTIVE

HOW MY PROCESS HAS CHANGED

This project reinforced a systems‑first approach grounded in observed behavior and usage constraints. I now begin system design by understanding what users are trying to accomplish and how they naturally engage with AI, then shape the interaction architecture and patterns around that, rather than starting from predefined UI components.

WHAT COULD BE IMPROVED

Future iterations would extend the system into more complex AI workflows and define clearer guidance around agent behavior, tone, and interaction consistency as AI capabilities evolve. I would also formalize governance for introducing new AI patterns, so the system can evolve without fragmenting.

Emma Blackwell

Product Researcher & Designer

2026 Designed by Emma Blackwell. All rights reserved.

Emma Blackwell

Product Researcher & Designer

2026 Designed by Emma Blackwell.
All rights reserved.