Password Protected

This case study is available on request.

Incorrect password. Contact avproductdesign@gmail.com
Case Study · 01

Making AI intelligence
feel human

Nimbus Intelligence
Freelance UX Designer
Enterprise Onboarding · Workflow Builder
2025 – 2026
Background

An AI platform built for enterprise strategy teams

Nimbus Intelligence is building a real-time decision intelligence platform — helping product companies validate ideas against live market signals before they build. An always-on intelligence engine for the modern enterprise.

I joined as a freelance UX designer working directly with the founding team. My mandate: take a genuinely powerful AI engine and make it usable for enterprise clients from the moment they sign up.

AI-powered market intelligence that fuses external signals with internal company data to guide product decisions in real time.

Enterprise teams across all levels — C-suite needing board-level insight through to product managers running day-to-day research workflows.

Enterprise clients at $50–200K/month need immediate value. A confusing onboarding or unconfigurable workflow is a churned account.

Embedded with founders — designing, iterating, and shipping in real time. No design team, no lengthy process. Fast, direct, high-trust collaboration.

The Challenge

Powerful infrastructure, zero usability

Nimbus had built something genuinely sophisticated — a perception engine processing millions of signals per minute, identifying threats, surfacing opportunities, and routing insights automatically.

The problem: none of that power was accessible to a non-technical enterprise buyer. Two distinct design challenges emerged.

Challenge 1 — Enterprise Onboarding

Clients needed to load their entire company context before seeing any value. Products, competitors, strategic priorities, roadmap — a massive data-entry problem that risked feeling like homework rather than a product.

Challenge 2 — Workflow Configuration

The intelligence engine ran on configurable AI pipelines. Powerful in theory — but presenting it as a technical interface would kill adoption with non-engineering buyers.

Work · Part One

Rethinking enterprise onboarding

The key insight: don't ask enterprise clients to fill out forms — use AI to do the work for them, then let them validate. By scraping the company's website on arrival, Nimbus pre-fills the entire profile automatically. Users confirm rather than input.

A real-time "Context Quality" score in the sidebar made the AI visibly smarter with each step — turning setup into something that rewarded completion.

1
URL entry — AI analysis begins
Enter company URL → AI immediately analyzes the site to kickstart setup
2
AI pre-filled company profile
Company profile pre-filled by AI — users validate, not type. Context Quality updates live.
3
Strategic themes with priority ranking
Strategic priorities selected and drag-ranked — AI tailors signal feeds accordingly
4
Intelligence activation — all modules ready
All four intelligence modules activated — unambiguous completion, ready to enter the workspace
Work · Part Two

A workflow builder for non-engineers

The workflow builder needed to give enterprise users genuine control over complex AI pipelines — without requiring engineering knowledge. The answer: a node-based canvas with three distinct panels working together.

Full workflow canvas overview
Full pipeline view — Competitor Mentions → Entity Recognition → Competitive Sentiment → Risk Analysis Subflow → Threat Alert → Executive Alert

Each panel serves a different purpose. The component library lets users build pipelines without writing code. The settings panel configures workflow-level behavior. The component panel exposes per-node AI parameters in plain language.

Smart components library and workflow settings
Smart Components
Workflow Settings
Left: drag-and-drop component library — Input, Processing, Decision, Action. Right: workflow-level configuration — scheduling, notifications, team routing.
Component settings panel — per-node configuration
Component Settings
Right: per-node configuration — competitor tracking with AI smart suggestions, team assignment, and notification channel controls.

The right settings panel is contextual — clicking any node on the canvas switches it from workflow-level settings to that component's specific configuration. Same panel, different context. No modal interruptions, no loss of orientation in the overall flow.

Design Decisions

Four choices that defined the product

01

AI pre-fill, human validation

Instead of blank forms, AI populates fields from website analysis. Users confirm rather than create — dramatically reducing friction while keeping humans in control of accuracy.

02

Context Quality as live feedback

A real-time completion score showed users how inputs improved the system's intelligence — turning setup into something that visibly rewarded effort and made the AI feel responsive.

03

Confidence threshold as slider

Rather than raw probability values, AI parameters were surfaced as labeled sliders. Technical control made human-readable without dumbing it down.

04

Node status always visible

Every pipeline component shows live status, throughput, and accuracy. The AI's behavior is legible at a glance — essential for building enterprise trust in automated systems.

Outcome

Infrastructure made human

Nimbus moved from a powerful but inaccessible AI engine to a product enterprise clients could configure, understand, and trust — without a technical onboarding call.

The onboarding flow eliminated manual data entry. The workflow builder gave non-engineering buyers genuine control over complex AI pipelines. Both shipped in tight collaboration with the founding team, iterating in real time.

Enterprise Onboarding AI Workflow Builder Progressive Disclosure Multi-stakeholder UX Founding Team Collaboration Fast Iteration 0 → 1 Product Design