0->1 Growth Systems for AI Activation + Monetization

Butternut's advanced AI website builder struggled with complex interactions and a confusing structure, preventing non-technical users from effectively creating their desired websites. Recognizing this as a critical barrier to market growth and user adoption, I led the transformation of the application, focusing on intuitive AI interaction design and a scalable design system. This initiative significantly improved user satisfaction and efficiency, boosting post-task satisfaction by 40% and making AI website creation accessible to a broader market segment.

ROLE

Lead Product Designer

TEAM

As the first design hire, I worked directly with the Founder and CEO, partnered with Product, and collaborated with Engineering.

KEY OUTCOMES

→ Trial-to-paid conversion increased from 8% to 13%.

• The business was freemium: generate free, pay to publish. I moved the paywall from early in editor to after first edit + preview. Resulted in a 62% relative lift in six weeks.

→ Onboarding completion increased by 32%.

• We replaced the blank prompt box with templated starting points. Non-technical users were able to reach their first generated website faster because they had examples to edit instead of writing from scratch.

→ Established a consistent AI interaction language across all features.

• I built the design system 0->1 to solve a consistency problem: every AI feature used different input patterns, copy, and empty states, so users couldn't build a mental model for how to talk to the AI.

→ Product-wide patterns were established.

• The templated prompt framework became the standard approach for all AI features across the product. The research synthesis I created became the template the team uses for future studies.

THE PROBLEM

The Business Problem We Were Solving

When I joined Butternut, the company had built impressive AI technology but users were not converting to paid plans.

Three specific barriers were blocking growth:

First, we had an activation problem.

• Non-technical small business owners did not know how to prompt the AI effectively. The onboarding flow showed a blank text box that said “Describe your website.” This led to a 15% error rate, and most users dropped off before the AI ever generated their first website.

Second, we had a monetization problem.

• Users could generate and edit free, but hit the publishing paywall early in the editor, before finishing their site. They had a draft, but hadn’t made it theirs yet. This resulted in an 8% trial-to-paid conversion rate, which was below the 12% benchmark for SaaS products at the time.

Third, we had a consistency and trust problem.

• There was no design system in place. Every AI feature had a different input format, different language, and different empty states. Non-technical users couldn't learn how to talk to the AI because it never spoke consistently.

The goal was to prove that an AI website builder could successfully activate and monetize non-technical small business owners.

THE PROCESS

How I Approached The Work

Phase 1: Aligning CEO and PM on definition of success

We agreed that the two metrics that mattered were the percentage of users who generate their first website and the trial-to-paid conversion rate.

Phase 2: Workshop with CEO, PM, and Engineering

We created two hypotheses. Hypothesis one: Blank prompts cause user drop-off, so providing templates will increase completion. Hypothesis two: Moving publishing paywall from early edit to after first edit + mobile preview will increase conversion by aligning ask with ownership.

Phase 3: Testing Hypothesis

We tested these hypotheses with 15 non-technical small business owners. We measured error rates, the time it took to reach their first generated site, and the specific objections they raised at the paywall. The testing showed that templates reduced errors and that showing the website before the paywall reduced the “I’m not ready” objection.

Phase 4: Shipped New Onboarding & Paywall

Within six weeks, onboarding completion increased by 32% and trial-to-paid conversion increased from 8% to 13%.

Phase 5: Systematized The Wins

I turned the prompt templates into reusable components in a design library. I documented the research approach so the team could repeat it. This enabled the growth team to run experiments every week instead of every month.

WHAT WE SHIPPED

Change 1: Paywall Timing

Before

Users encountered the paywall while still fixing their first draft. They hadn’t completed an edit or seen how the site looked on mobile. 92% dropped off before ever previewing, without experiencing ownership. The trial-to-paid conversion rate was 8%.

After

I redesigned the flow to let users finish before asking them to pay.

New flow: Generate for free → Guided first edit → Mobile preview with “Your site is ready” confirmation → Paywall “Upgrade to publish and go live.” The trial-to-paid conversion rate increased to 13%.

I led the design for the new flow. I partnered with the PM to plan and run an A/B test to validate the timing change.

Change 2: AI Interaction Design

Before

The onboarding flow showed a blank prompt that said “Describe your website.” Users did not know what information to include or how detailed to be. This caused a 15% error rate, and 68% of users did not complete onboarding.

After

The onboarding flow said “Pick a starting point” and showed three templated prompts with live previews of the types of websites they could build. Users selected a template and edited the content instead of writing from scratch. Onboarding completion increased by 32%.

I led the user research with 15 participants. I designed the templated prompt system and the interaction patterns for editing.

Change 3: Design System For AI Consistency and Trust

Before

No shared components or interaction rules. Every AI tool looked and behaved differently. Users had to re-learn how to prompt for each feature, which eroded trust and prevented adoption.

After

I initiated and built the design system 0->1, including our core AI patterns, onboarding steps, and paywall components. I defined a consistent language for how the AI asks, confirms, and handles errors, so users could build a mental model and trust the system.

I initiated and built the design system from scratch. I onboarded the junior designer and the engineering team to the system.

IMPACT

Product Impact:

• The paywall timing change is still how Butternut monetizes today. The templated prompt framework is now used across all AI features in the product, not just onboarding.

Organizational Impact:

• The research synthesis document I created became the template that PMs and designers use for future studies. The component library I built became the single source of truth for the design and engineering teams.

What I Learned:

• For AI products, you need to prove the magic before you ask for the credit card. Activation means removing cognitive load so users can reach their first win without having to guess what to type.

Acknowledgements

Acknowledgements This transformation was a collaborative effort. My sincere gratitude and congratulations go out to every talented individual involved in bringing this vision to life.