How I Designed An AI UX Assistant That Knows What You Need, Before You Ask

NUVUN is a lean, AI-powered, Zero UI research and design assistant built to anticipate needs, connect insights across projects, and orchestrate tools in real time. It works across voice, text, image, and video, and delivers actionable outputs directly where teams work, without adding new platforms or learning curves.

Timeline

Ongoing development

Role

Product Strategist & Product Designer

Method

Journey mapping + A/B testing + change management

Team

1 developer, 1 AI engineer, 1 UI Designer

Business Problem

As teams grow, they lose UX knowledge, repeating studies and relying on assumptions. Traditional research is too slow and expensive to scale. This leads to low adoption features and wasted resources. NUVUN automates and structures research to make it accessible within real team workflows.

What We Got Wrong

Failed hypothesis #1

AI could fully replace research expertise in small teams.

Failed hypothesis #2

Only trained researchers could generate meaningful insights.

Failed hypothesis #3

Advanced proactive AI should be built first for maximum value.

Controversial Decisions

The “No New Tool” Decision

We decided not to build a standalone platform and instead integrate NUVUN into tools teams were already using, like Slack, Notion and Google Meet.

Why this was controversial

Building a full-featured app would offer more control and visibility. But we prioritized zero learning curve and minimal disruption, even if it meant sacrificing some UI flexibility.

The Transparency-by-Default Rule

Instead of keeping AI reasoning hidden, NUVUN always shows the source data and confidence level for every insight.

Why this was controversial

Some feared it would overwhelm users or expose AI limitations. We believed transparency was essential to build trust, especially when insights could influence high-stakes product decisions.

V1: Lean Research Automation

• Delivers quick wins through a PRD Document, no learning curve or extra tools.

• Designed for full-cycle research across multiple departments at launch.

• Envisioned a separate platform with its own interface.

V2: Memory-Based Knowledge Assistant

• Starts with lightweight automation and conditional logic to deliver quick results.

• Acts as a contextual memory layer that helps teams scale their strategic thinking over time.

• Works invisibly within tools teams already use, extending their workflows without disruption, reducing cognitive load.

Comparison

vs. Hiring UX Researcher

More consistent, no sick days, scales infinitely, preserves knowledge forever

vs. Generic AI (ChatGPT, Claude)

Domain expertise, structured memory, integrated workflow, learns from your specific context

vs. Documentation Systems (Notion)

Active intelligence vs. passive storage, proactively suggests relevant insights.

Research Cycle Reduction

Reduce research cycle time through more intelligent flows

x3

User Data Touchpoints

Multiply by 3 user data touch points per sprint.

Cross-Functional Participation

Increase team participation in research in non-UX roles.

Where we're going

Proactive Insight Delivery:

Anticipate research needs before they’re requested using real-time pattern detection.

Multimodal Research:

Process text, voice, image, and video in a single research flow. 

Context Persistence:

Connect new findings to historical insights across projects. 

Tool Orchestration:

Trigger actions in Calendly, Jira, Figma, Google Slides, and more.

Transparent Reasoning:

Show data sources and confidence levels for every insight.