MirrorWorx — Designing a Personalized AI-Assisted Product System

Turning owned products, verified client context, and AI-assisted reasoning into usable, step-by-step routines.

Built a working client experience that connects persistent inventory and verified context to AI-assisted recipe composition, human review, progressive disclosure, and exact execution guidance.

Role

Founder / Product Designer

Company

MirrorWorx

Capabilities

AI-Assisted Product Design, Personalization, Information Architecture, Structured Data, Human-AI Interaction, Cross-Device UX, WordPress Implementation

01 · Context

The Challenge

Most beauty and personal-presentation recommendations begin with a generic question: what should someone buy or use? MirrorWorx starts somewhere more useful: what does this client already own, what is verified about those products, what are they trying to accomplish, and how should that context change the recommendation?

I designed MirrorWorx as a personalized product system that turns persistent inventory, verified client context, and AI-assisted reasoning into executable routines. The interface does not stop at a recommendation; it translates the recommendation into exact products, shades, tools, quantities, placement, and application methods that can be followed beside a mirror on a phone or iPad.

Changing the organizing model

The product reframe moved MirrorWorx from advice about products to a system that can reason from a client’s actual inventory and context, then expose only the part of the routine the client needs in the moment.

BeforeGeneric recommendations and monolithic instructionsPrompt → suggestion → user interprets what to do

AfterContext-grounded recipes with executable guidanceGoal + inventory + verified context → assisted draft → review → Full / Complexion / Eyes / Lips → step-by-step execution

What I Owned

Defined the product concept, information architecture, and client context model

Structured owned products, shades, tools, preferences, recipe context, verification status, and finished looks as connected product information

Designed the client dashboard and LOOK recipe experience

Designed progressive recipe views for Full Recipe, Complexion, Eyes, and Lips

Defined step-level fields for function, placement, quantity, application method, exact product or shade, and exact owned tool

Established verified versus provisional information and a Ready-state review gate for assisted recipes

Designed the experience for phone and iPad use beside a mirror and during filming

Implemented the working experience in WordPress and iterated directly in the product

How the Direction Changed

Persistent context

Move from one-time prompts to reusable inventory, product identity, shade, formula, tool ownership, preferences, and prior client context.

Trust before generation

Separate provisional information from verified facts and require human review before an assisted recipe becomes Ready.

Progressive disclosure

Preserve one underlying LOOK while allowing Full Recipe, Complexion, Eyes, and Lips views for the task at hand.

Executable guidance

Encode purpose, placement, quantity, formula-appropriate method, exact product or shade, and exact owned tool at the step level.

The resulting loop connects client goal and context to an assisted draft, constrains recommendations with verified inventory, uses human review to govern readiness, and delivers a focused routine that can be reused across devices and contexts.

Selected Designs

Finished experience: Quick Polished — Warm Gold connects verified context and owned products to an executable routine and visible result. — Finished experience: Quick Polished u2014 Warm Gold connects verified context and owned products to an executable routine and visible result.

Client Dashboard: personalized LOOK recipes are presented as usable client experiences rather than generic content.

Inventory-grounded recommendation: the recipe shows the exact products and shades already in the client’s inventory.

Eyes view: progressive disclosure isolates the eye products and steps needed in the moment while remaining part of the same LOOK.

Lips view: the same architecture scales to another recipe domain without creating a disconnected experience.

Executable eye guidance: each step combines purpose, quantity, application method, exact shade, exact product, and the client’s owned tool.

Executable lip guidance: product and tool references stay attached to the specific action the client needs to perform.

What I Shipped

Built a working client experience that connects persistent inventory and verified context to AI-assisted recipe composition, human review, progressive disclosure, and exact execution guidance.

Personalized Client Dashboard — surfaces LOOK recipes built from confirmed owned inventory

Context-Grounded Recipe System — connects verified client context and persistent inventory to assisted composition

Progressive Recipe Views — Full Recipe / Complexion / Eyes / Lips from one underlying experience

Executable Step Model — placement, quantity, method, product, shade, and owned tool remain attached to each action

Human Review & Verification — prevents provisional AI-assisted information from automatically becoming authoritative

Cross-Device Experience — practical guidance designed for desktop management and iPad or phone use beside the mirror

MirrorWorx is an independently developed product in active development. This case study documents implemented product capability and design decisions; it does not claim mature-market adoption metrics.

Why this pattern matters

The underlying problem is not beauty-specific; it is context management. MirrorWorx explores how an AI-assisted product can maintain structured knowledge about a user, distinguish verified facts from provisional information, constrain recommendations with that context, and translate reasoning into actions a person can actually perform.