Ai Assistant

Human-Centred Product Discovery Through Prompt-Based Design

Role: Lead Product Designer
Context: B2C | AI Integration | Modular Product E-commerce

🧠 The Challenge

As part of 3D Cloud’s ambition to stay at the forefront of technological innovation, we sought to explore where AI could meaningfully support customer experience.

A key challenge lay in the complexity of modular product configuration — users were overwhelmed by choices, resulting in high drop-off rates and lengthy decision-making journeys. With limited in-house AI experience, our strategy was to pilot a client-facing AI tool to tackle a clear UX bottleneck.

⚽️ Strategic Goal

To position 3D Cloud as a forward-thinking leader in modular commerce by integrating accessible, conversational AI into the product configuration experience—reducing cognitive load, accelerating decision-making, and increasing customer satisfaction. This initiative supports a broader business objective to drive product discoverability, improve conversion rates, and differentiate our digital offering in a competitive e-commerce landscape.

🎯 Design Objectives

Simplify the product discovery process via a conversational AI interface.

Enable text and image-based input to guide product configuration intuitively.

Drive AI adoption without alienating users unfamiliar or hesitant about the technology.

Ensure accessibility and mobile readiness from day one.

Just for context

Here is a live example of the modular configurator application deployed for Design with in reach (taken 10.01.2025). A US home and office furnishing customer providing direct customer access via their e-commerce platform.

✍🏼 Wireframes

🧭 Approach & Methods

🔍 Research & Strategy

  • Customer persona creation based on behavioural data

  • Analytics review to identify PDP abandonment points

  • Client workshops to define AI boundaries, tone, and product relevance

  • Competitor scan across emerging AI interfaces (e.g. Shopify Magic, Adobe Firefly)

  • Journey map + sitemap update for AI integration touchpoints

🔧 Design & Prototyping

  • Conversational UX flow mapping for AI prompts and responses

  • Component audit and scalable design updates

  • Rapid prototyping using scenario-based product queries

  • User testing (video walk-throughs, live tool trials, in-product surveys)

  • Cross-functional coordination with PMs, BAs, QA, and AI data trainers

📤 Takeaways

  • Some users hesitated to try the AI, but once they did, they wanted it in every shopping experience.

  • Labeling the tool as “AI” slightly reduced first-time use, but increased return visits.

  • Users praised the tool’s speed and the natural, free-flowing interaction.

  • The tool worked seamlessly on mobile — no changes needed.

  • Voice input and image uploads gave the tool a unique advantage over other configurators.

🎨 Final Design

Simple user flow

✨ Key Design Solutions

Impact

Area

Solution

Prompt-based input for conversational setup

AI Interface

+60% faster product creation

Image + voice note inputs to guide AI

Media Inputs

Differentiates experience from competitors

Smart, context-aware product suggestions

UI Feedback

↑ Product exploration depth

Instructional video to build trust & drive adoption

Content

+73% NPS from AI users

Seamless experience across devices

Accessibility

Fully responsive, mobile-ready design

📈 Outcomes & Business Impact

Result

Metric

🛒 Add to cart rate

↑ 43%

🤖 AI adoption rate

86%

❌ PDP drop-off rate

↓ 17%

↑ 12% more options viewed

🧠 Avg. configuration breadth

+73% of users more likely to recommend brand after using AI tool

📣 Brand advocacy

🔄 What I Learned

🛬 Before Launch

  • AI models are only as powerful as the data they ingest — training quality dramatically affects output.

  • Internal collaboration was essential: design, QA and data teams co-owned model training to speed up results.

🚀 After Launch

  • Users created configurations 60% faster using AI vs. menus.

  • Fear of AI was a barrier for some — however, video tutorials helped improve uptake.

  • Mentioning “AI” in the tool lowered initial interaction but increased return visits.

  • Users described the tool as “fast”, “natural”, and “surprisingly fun”.

🧠 Strategic Takeaway

AI-driven tools must feel intuitive, not futuristic. Simplicity, guidance, and trust-building elements (like “how it works” videos) were essential for user engagement. If repeated, I’d recommend:

  • Earlier pilot builds with messaging that this is a “learning tool” in beta — encouraging feedback and lowering expectations.

  • Staggered onboarding with optional AI tutorials or voice walkthroughs to widen adoption.