CASE STUDY

Reimagining Lightcast with an AI-powered composable UI.

GenAI

Composable UI

Conversational UI

Heuristic evaluation

Data-viz

Recruiting & Talent

At Lightcast, I designed the conceptual framework for an AI-first report composer. The UI begins not with filters or data fields, but intent. The user simply states their goal, and the system surfaces modular data blocks based on relevance. From there, users can drag, combine, or refine those blocks into a tailored report.

My team is building an AI-powered, blockchain solution that composes the supply chain with AI, and processes air freight bookings via the novel smart-contract protocol, OXP.

Unlike traditional tokens and NFTs, OXP is an intelligent orchestration protocol that tokenizes outcomes, allowing it to: a) track shipments through completion, b) enforce trustless transparency, and c) pay vendors automatically via escrow release upon the completion of service-milestones.

ROLE

Lead UX/UI Designer

AI strategist

ROLE

Lead UX/UI Designer

AI strategist

TIMELINE

3 months

Mar. 2023–Aug. 2023

TIMELINE

3 months

Mar. 2023–Aug. 2023

SCOPE

Heuristic evaluation, AI strategy, all design concepts

SCOPE

Heuristic evaluation, AI strategy, all design concepts

TEAM

Head of Product, Product Manager, 2 AI/ML Engineers, 1 Data Engineer

TEAM

Head of Product, Product Manager, 2 AI/ML Engineers, 1 Data Engineer

THE PROBLEM

Solving for high-friction on the path to data discovery and report building, for recruiters assessing the candidate marketplace

Recruiters using Lightcast struggled to generate insights quickly due to a labyrinthine report experience. They were required to select the right report from a library of dozens, then navigate dense filter menus and complex form fields—only to be overwhelmed by dense reports that included much more than what they asked for.

Recruiters using Lightcast struggled to generate insights quickly due to a labyrinthine report experience. They were required to select the right report from a library of dozens, then navigate dense filter menus and complex form fields—only to be overwhelmed by dense reports that included much more than what they asked for.

The existing flow

Recruiters started by choosing from a large report library, even when they only knew the hiring question or market problem they were trying to solve.

To get useful data, users had to navigate dense filters, rigid forms, and field logic before they could understand what was available.

Reports returned broad charts and tables, but buried the most relevant signals inside static, overloaded data views.

Even when users found valuable data, they lacked a simple way to save, organize, edit, and turn those findings into a clear report.

UI PREVIEW

By using AI to source data directly, Lightcast was able to salvage 200 of 250 at-risk account who piloted the redesign. This UI from early 2023 was ahead of its time — featuring an outcome-oriented conversational LLM that surfaced data visualizations and insights, which could then be dragged into the composable Report Builder.

UP FIRST: PRODUCT EVALUATION

I began by assessing their product against the 10 UX heuristics and 101 psychological factors of user experience. I discovered 120+ usability issues and synthesized findings into 4 UX themes that defined a cohesive North Star product vision forward.

Issues ranged from low-hanging fruit to critical systemic issues rooted in the information architecture. The initial deliverable was a slide deck in Figma, outline the usability issues, architectural problems, a rating, and a recommended path forward.

The anatomy of a slide: each slide aggregated findings for a single screen or element, clearly explained the problem, itemized transgressions of heuristics and psychological principals, a proposed solution hypothesis, and listed UX themes and rated impact in the top right.

The original product's flow had substantial friction

The previous product forced users to select a report

Dense forms presented as substantial friction to customize the report

Reports presented broad data that buried the user's desired insight

THE SOLUTION

AI Search: I synthesized the heuristic evaluation and proposed an AI strategy that flipped the user model: instead of navigating through circuitous tree-navigation to data, the user would summon the data directly by expressing what they’re trying to achieve in natural language.

Report Building: AI then surfaces relevant data blocks as insights, contextual analyses, and data visualizations, to be dragged as modular and composable data blocks into the final report.

UX THEMES

Four UX themes became the pillars of a cohesive AI product redesign

The original UI forced users to match their search intent to the right report. The redesign shifted the conversation from, "select a report" to “what are you trying to achieve?”, using LLMs to let users summon relevant data through intent.

The original UI forced users to match their search intent to the right report. The redesign shifted the conversation from, "select a report" to “what are you trying to achieve?”, using LLMs to let users summon relevant data through intent.

I created custom illustrations to help stakeholders visualize the complex themes of the redesign. Keep in mind, the work here took place in 2023, before AI image generation. I have since restyled the original illustrations with AI. This visual approach is not typical of my traditional Design Thinking process, but I used it here as a means to present and get buy-in for the North Star concept.

Theme #1

Outcome-oriented IA

I recognized early on that AI inverts the relationship between users and data. The core paradigm of AI product design therefore, outcome oriented information architecture, allows users summon the data to them, where instead of forcing them to navigate circuitous tree structures, they simply express the outcome they want to achieve.

Outcome-oriented IA is depicted as nested outcomes, with micro-outcomes nested within sub-outcomes within macro-outcomes.

Theme #2

Modular composability

In build workflows, users can select the “gems” from AI-generated results and seamlessly assemble them into a final report. Insight becomes composable—built progressively as users construct and cherry pick top deliverables within the report workflow.

We use AI to illustrate our thinking and hang artifacts on the wall of the office.

Theme #3

Semantic data-mapping

Instead of static categories, data is connected through meaning. Users move fluidly between concepts—mirroring how they think, not how the system is organized. This theme creates 'semantic neighbors' of both insights and outcomes to accelerate user momentum. In the UI, these surfaced as suggested follow-up prompts.

The conveyer belt shows search outcomes being converted into data visualizations that are added to a report to illustrate modular composability.

Theme #4

Untapped value

User research revealed that recruiters are unable to craft their reports in-product, and are forced to build it in third party apps as a workaround—a missed value-add opportunity for the business. Moreover, users frequent the same 1-2 reports, and do not venture out to explore unfamiliar reports, and the underlying value of those reports remains untapped. In a sense, users were unable to access the full value of the data in the platform. This was a proposal to add in-app reporting to the platform.

We use AI to illustrate our thinking and hang artifacts on the wall of the office.

DEFINE & IDEATE

Semantic data mapping

AI identifies the user’s intended outcome. Semantic mapping connects that outcome to adjacent outcomes in the form of related prompts and reports, streamlining user momentum as they work to build a complete report. Semantic data mapping defined how one user intent could open a network of related insights.