03_SneakPeek_Givebutter_Growth_Design

10 min read

Overview

Clint, a Staff Product Designer on the Growth team at Givebutter, walks through his AI-native growth design workflow with host Jay. The core job of a growth designer, he argues, is “professionally noticing friction” — not waiting for a PRD but actively hunting problems. His process chains together specialized tools: Enterpret (voice-of-customer feedback AI) and Hex (behavioral/quantitative analytics) run in tandem, meeting context is captured by Granola, and Claude is used to turn meeting “nuggets” into structured prompts that interrogate Enterpret and generate opportunity areas. Concepts are rapidly prototyped with Claude Code (for novel, free-form ideas) and Magic Patterns (for refining real product flows against the actual Storybook design system), then landed in Figma as the source of truth. A 2FA and a “switcher intent” onboarding experiment show how this speed enables fast learning and pivoting — including discovering a wrong mental model (only ~17% of users actually self-identified as switchers). Mobbin serves as the “visual research companion” that feeds taste and reference screenshots into Claude and Magic Patterns.

Key Insights

  1. The growth designer’s core job is “professionally noticing friction” — not waiting for a PRD. Clint spends less time pushing pixels and more time understanding why users get stuck. He actively jumps into dashboards and user feedback rather than waiting for a PM checklist or a PRD to appear. (0:03, 4:04)

  2. Qualitative and quantitative tools are used in tandem. Enterpret gives the “emotional texture” behind a problem — how users feel — while Hex shows what users actually do. Clint merges both sides (qualitative + quantitative) into a single friction-hunting picture before deciding which problems to pursue. (4:46, 5:00)

  3. Meeting context (Granola) is the raw material for problem discovery. Granola sits on his desktop capturing notes from all conversations. He uses it as a searchable reference (“what did so-and-so say about friction at sign-up?”) and pulls those nuggets into Claude to generate prompts to query Enterpret. (6:07, 10:45)

  4. Claude converts rambling into structured prompts for other tools. Rather than writing prompts by hand, Clint pastes conversation context into Claude and asks it to produce grouped prompts by theme, which he then feeds into Enterpret with a fixed “framework” attached (summarize patterns, focus on missed opportunities, draft a brainstorming doc organized by opportunity area / user stories / citations / experiment ideas). (7:25, 11:33, 12:06)

  5. Voice of the customer literally enters the design process. Enterpret surfaces real Intercom feedback and Gong call recordings, so Clint can read users’ words and listen to their frustration/excitement while designing — and later cite real user quotes to justify design direction. (13:05, 13:35)

  6. AI-generated experiment ideas become a pipeline. Enterpret’s output includes concrete experiment ideas (e.g., a “Post-campaign CRM trigger” modal). Clint takes promising ones, spins them up in Claude, and pushes them forward — but he also banks unpursued ideas in a “bank of ideas” he can revisit. (14:05, 15:51)

  7. Prototyping tools are split by intent: refine vs. invent. Magic Patterns is for refining real product flows with prompts (tied directly to the Storybook design system, so output uses actual production components, tokens, icons); Claude Code is for prototyping things that don’t exist yet, freeing him from design-system constraints to explore ideas in a free-flowing direction. (29:20)

  8. Figma remains the source of truth and team convergence point. One Figma file per quarter (a practice adopted from Jay’s interview with the Firefox team) is where everything lands. Once a concept is “vibing,” Clint asks Claude to send every prototype screen to Figma to prepare for handoff, because Figma is where conversations and real components/Storybook additions are tracked. (27:03, 28:01)

  9. Speed = learning faster, not just shipping faster. The switcher-intent onboarding experiment went from concept to working prototype to deployment-ready in 2 days. The team learned within ~2 weeks that its “switcher intent” mental model was wrong — only ~17% of users confirmed they were switching. The real lesson: the win was learning fast enough to pivot before over-committing, not proving a hypothesis right. (35:06, 36:25, 37:03, 49:49)

  10. The AI-native workflow has fundamentally changed what a designer’s day looks like. Clint notes his workflow changes roughly every 3 weeks (V0/Cursor → Magic Patterns → Claude, sometimes Figma, sometimes not) and that roughly half his job is now “aggressively vibe-checking robots” — robots produce brilliance sometimes and garbage other times, and both are useful because you learn from both. (50:27)

Actionable Techniques

  1. Adopt “friction hunting” as a weekly ritual. Every week, deliberately search the product for where users get stuck — onboarding, drop-off points, activation — using both qualitative and quantitative data, instead of waiting for a PRD or checklist. (4:36)

  2. Run Enterpret + Hex in tandem. Use a feedback-intelligence tool for the emotional/qualitative layer (what users feel/say) and an analytics tool for the behavioral layer (what users actually do), then merge both into one problem picture. (4:46, 5:00)

  3. Use Granola (or similar meeting-capture) as a searchable memory. Let it sit on your desktop recording meetings, then ping it with questions to retrieve exact statements (“what did the PM say about friction at sign-up?”). (6:07, 10:45)

  4. Have Claude turn meeting nuggets into prompts for your feedback tool. Paste conversation context into Claude, ask it to generate theme-grouped prompts, then drop one or two into Enterpret. (7:25, 11:33)

  5. Attach a fixed “framework” to every prompt. Alongside the question, ask the tool to: summarize feedback and patterns, focus on missed opportunities, draft a brainstorming document organized by opportunity area / user stories / citations, and propose experiment ideas. (12:06)

  6. Read and listen to actual customer feedback while designing. Drill into the raw Intercom threads and Gong recordings behind the AI summary so the real voice of the customer (frustrations, excitement, exact wording) is driving your decisions, and collect real user quotes to justify design direction. (13:05)

  7. Keep a “bank of ideas” for off-roadmap findings. When you stumble on a high-value problem that isn’t scheduled, start a Claude session to explore it, pin it, and return later — build a repository of ideas you can pick back up at any time. (15:51)

  8. Communicate findings with a quick Loom + Slack thread. When you learn something, record a short Loom (keep it under 5 minutes, set default playback to ~1.5–1.75x to respect the team’s time) and drop it into Slack with links to clickable prototypes, letting the idea “percolate” and gather async comments/feedback. (39:15, 40:29)

  9. Build multi-state prototypes to document flows in one place. Instead of separate prototype links per flow, build one prototype with navigation controls to switch between states (e.g., first visit, partial progress, completed, returning user) plus a small “growth team test” badge so stakeholders know who to ask about it. (24:26, 25:21)

  10. Choose your prototyping tool by refine-vs-invent. Use Magic Patterns (with your design system wired in) to iterate on real flows and production components; use Claude Code for novel ideas outside design-system constraints. (29:20)

  11. Feed taste/reference screenshots directly into AI tools. Use Mobbin to search specific patterns (e.g., security pins and 2FA), then screen-cap or download references and paste them into Claude or Magic Patterns (“I want to try something like this — here’s a screenshot”). This makes inspiration a frictionless, searchable part of the process. (45:02)

  12. Land everything in Figma as source of truth. When a concept is approved, ask Claude to push every prototype screen into Figma for documentation and handoff, and use one shared Figma file per quarter to keep the whole team on the same page. (27:03, 28:01)

  13. Design prototypes to look and feel real so you can test in-product quickly. The 2FA and switcher experiments were built as realistic prototypes, letting the team get fast feedback and test variants (e.g., ABC tests of three onboarding survey versions) within days — enabling fast failure and pivoting. (35:06, 36:25, 49:49)

  14. Prototype with “one-click” frictionless patterns. When you must ask users something, prefer a single one-check interaction over a drop-down plus yes/no — it requires almost no work and still returns signal. (0:52, 42:27)

Tools & Skills Mentioned

  • Enterpret — customer feedback AI; aggregates Intercom, Gong, G2, and mobile/web feedback; Wisdom tool for prompting questions over the voice of customer; outputs opportunity areas, user stories, citations, and experiment ideas
  • Hex — behavioral/quantitative analytics (“what users actually do”); median sequence of actions analysis; talk-to-data Slack channel (“talk to data”) for chatting with data from within Slack
  • Granola — desktop AI meeting-notes app; searchable meeting memory; used for retrieving exact statements (not currently MCP-connected in his setup)
  • Claude / Claude Code — used to convert rambling into structured prompts, to generate Enterpret queries, and to prototype novel/exploratory ideas outside the design system
  • Monologue — voice-to-AI input tool (hold home key, speak what you’re looking for); stats dashboard; switched from WhisperFlow ~2 months prior
  • Whisper Flow — earlier voice-input tool he used before Monologue
  • Magic Patterns — AI prototyping/refining tool; connected to the Givebutter Storybook design system (typography, color tokens, icons); Canvas tool for organizing exploration history
  • Cursor / V0 — earlier prototyping tools he started with (before Magic Patterns)
  • Figma — source of truth and team convergence; one file per quarter; place where designs are documented and handed off; real components & Storybook additions tracked
  • Mobbin — UI pattern reference library / “visual research” companion; used to search specific patterns and gather screenshots for inspiration
  • Storybook — design system source (components wired into Magic Patterns)
  • Loom — async video updates for communicating explorations and gathering feedback
  • Granola (meeting capture) — raw material for problem discovery
  • GuideStar — verified nonprofit database used to tailor onboarding survey questions per user
  • Slack — collaboration surface for sharing Looms, prototypes, and feedback threads

Quotes Worth Keeping

  • “A lot of what I do as a growth designer is just professionally noticing friction in the product.” (0:03)
  • “I’m not waiting for a PRD to show up. I’m actively jumping into these dashboards and the feedback that we’re getting from our users.” (0:26, 10:27)
  • “I spend way less time pushing pixels and more time trying to understand why people are getting stuck.” (4:26)
  • “Enterpret is one of those where I can hear the emotional texture behind the problem… Hex is more of what users are actually do.” (4:46, 5:00)
  • “I literally have the voice of the customer in my head as I’m designing through these things.” (13:26)
  • “The win here wasn’t like proving ourselves right. The win was learning fast enough to pivot before we over-committed to the wrong direction.” (38:16)
  • “We realized that the issue wasn’t in the UI, but it was our mental model.” (37:41)
  • “Speed is not just shipping things faster, it’s learning faster.” (49:51)
  • “Half my job now is aggressively vibe-checking robots… Sometimes they give you brilliance and sometimes they give you absolute garbage… Both of those are useful because you’re learning from that.” (50:54)
  • “It’s still all about helping people to do meaningful work.” (51:13)
  • “Figma is still very much a part of our process as a team to really get everyone on the same page.” (28:37)
  • “It’s become this very easy, frictionless part of my process where I can easily go and search something, get unstuck, or just get inspired by things that other people are doing.” (46:17)

Graph View

Backlinks