Chetan J
Figma · Product Improvement Case Study
Product Improvement Case Study

Rebuilding Trust
After UI3

How a major redesign and AI data-sharing controversy tested trust with Figma's power users, and a PM framework for earning it back through progressive migration and transparent AI.

$1.06B
FY2025 Revenue
+41% YoY, first billion-dollar year
~2–3★
Trustpilot Rating
Rated 'Poor': UI3 & pricing backlash
Dominant
Designer Market Share
Default tool for pro UI/UX design
136%
Net Dollar Retention
Highest in 10 quarters
01 — Comprehend

The Problem

Figma began rolling out its UI3 redesign around Config 2024 and fully migrated users on April 30, 2025, effectively retiring the previous interface. At the same time, Figma introduced new AI features and a data-sharing policy that, by default, enrolled Starter and Professional teams to let their content be used for AI training, while Organization and Enterprise customers were opted out by default. These changes triggered intense backlash across forums and social media, and later became part of the context for a proposed class-action lawsuit alleging that Figma misused customer designs to train its AI models without proper consent.

"UI3 is, without a doubt, the most unfriendly design possible. It has made my workflow painfully slow and incredibly unproductive."

Representative composite, based on Figma Forum feedback threads

"UI3 feels like Illustrator for babies. The interface is so horrifyingly oversimplified that it makes Microsoft Paint look like feature-rich professional software."

Representative composite, based on Figma Forum and Reddit posts

Quotes are anonymized composites derived from public Figma Forum and Reddit posts, edited for brevity and clarity.

Trust Erosion Taxonomy

Forced UI3 Migration (Apr 30, 2025)
UI2 fallback permanently removed. No opt-out. Power users report noticeable productivity drops and friction.
All users
Critical
AI Training Default Opt-In
Individual/small team designs opted in to AI training by default. Organization/Enterprise tiers were not auto-enrolled. Opt-out buried in team settings.
Free + Pro users
Critical
Make Designs → Apple Weather Clone
AI feature generated near-replica of Apple's Weather app. Feature pulled within days of Config 2024 launch.
Public incident
High
Proposed Class-Action Lawsuit (Nov 2025)
Khan v. Figma: plaintiffs seek substantial damages and injunctive relief over alleged secret IP harvesting for AI training.
Pending litigation
Critical
Dev Mode Paywall ($25–35/seat/mo)
Previously free feature became paid. Teams reported near-doubling of annual spend. 'Very Adobe of you.'
Org + Enterprise
Medium
Bottom Toolbar: Non-Configurable
#1 unresolved UI3 complaint through March 2026. Conflicts with OS docks, overlaps work.
All users
High
02 — Research

Business
Context

Figma's situation created a clear tension: the product continues to grow as a core design platform, but some of its most engaged users voiced strong frustration with how major changes were rolled out. The metrics that matter to investors (revenue, retention, expansion) can diverge from the sentiment that matters to the community using the product every day.

IPO & Market Cap

IPO July 31, 2025 at $33/share on NYSE (FIG). Closed first day at $115.50, implying ~$68B market cap, before trading well below that peak by early 2026. $1B breakup fee from failed $20B Adobe acquisition (Dec 2023) padded the balance sheet.

Revenue & Retention

FY2025 revenue $1.056B (+41% YoY). Q4 2025 revenue $304M (+40% YoY). 136% net dollar retention, highest in 10 quarters. 95% of Fortune 500 companies use Figma. 1,405 customers at $100K+ ARR; 67 at $1M+.

Platform Expansion

Config 2025 doubled products from 4 to 8: Figma Make (AI code gen), Sites (web publishing), Draw (illustration), Buzz (marketing). Strong multi-product adoption driving expansion revenue. Strategy: 'design tool → product development platform.'

Competitive Landscape

Figma holds dominant share among pro UI/UX designers. Sketch at ~4.5% (Mac-only). Penpot (open source) has grown meaningfully on the back of Figma backlash. Adobe XD in maintenance mode. AI competitors: v0.dev, Lovable, Bolt.new, Google Stitch.

UI3 + AI Trust Erosion Timeline

Jun 26, 2024
Config 2024
UI3 announced. AI features debut. 'Make Designs' enters beta. Limited opt-in.
Jul 2, 2024
Apple Weather Clone
AI generates Apple Weather replica. Feature pulled. CEO admits can't confirm training data.
Aug 15, 2024
AI Opt-In Default
AI training on customer data begins. Org/Enterprise not auto-enrolled. Free + Pro users opted IN by default.
Oct 10, 2024
UI3 Available to All
Full rollout with opt-out toggle. Users can revert to UI2. Feedback overwhelmingly negative.
Apr 30, 2025
UI2 Killed
Forced migration. No revert option. Forum threads explode. Chrome extension built to override CSS.
Nov 21, 2025
Class-Action Filed
Khan v. Figma (N.D. Cal.): proposed class action alleges secret IP harvesting for AI training without consent.
03 — Solution

The Trust
Recovery

This case study explores two interconnected frameworks that could have reduced friction and preserved more trust during this transition. First, a Progressive Migration System that gives power users more agency over interface changes. Second, an AI Transparency Layer that makes data use and consent more visible, granular, and easy to manage.

01

Configurable Workspace

Toolbar position (top/bottom/side), panel docking (floating/fixed), and density presets (compact/default/spacious). Save per-user. Never force a single layout.

02

Gradual Feature Introduction

New UI features launch as opt-in toggles with clear before/after previews. Users adopt on their timeline. Deprecate legacy only after 80%+ voluntary adoption.

03

AI Consent Dashboard

Account-level (not team-level) AI data controls. Visual log of what data was used, when, and for which model. One-click opt-out for all AI training. Same default protections across all tiers.

04

Design Provenance Tracking

Every AI-generated element tagged with origin metadata: model version, prompt, training data scope. Exportable provenance certificate for IP protection.

05

AI Output Guardrails

Similarity detection against known design systems (iOS, Material, etc.) before output. Flagging when generated UI resembles existing products. User controls threshold sensitivity.

06

Community Feedback Loop

Public feature status board with voting. Monthly 'Design Council': top 20 power users by usage get direct access to product team. Forum threads never locked while active.

04 — Wireframes

Before → After

CURRENT: UI3 (FORCED)
▢ ◇ T ✎ ⊕⚠ FIXEDLayersFloating panelGap shows canvas✗ Boolean ops in overflow menu✗ Toolbar position not configurable✗ No density controlsPropertiesReorganizedPosition/sizelocation varies
PROPOSED: CONFIGURABLE WORKSPACE
▢ ◇ T ✎ ⊕ ∪ ∩ −✓ Position: configurableLayersDocked, no gap✓ Density: Compact / Default / Spacious✓ Boolean ops restored to toolbarAI Consent DashboardTraining: OFF · Provenance: ONAccount-level · One-click toggleNew Feature: Figma Draw⬤ Opt-in · Preview availablePropertiesStable layout
05 — Prioritize

RICE Scoring

FeatureReachImpactConfidenceEffortScore
Configurable workspace (toolbar + panels)All users (est.)395%2185.3
AI Consent Dashboard (account-level)All users (est.)390%2175.5
Gradual feature introduction (opt-in)All users (est.)290%2117.0
Design provenance trackingPaid teams (est.)280%328.8
AI output similarity guardrailsPaid teams (est.)270%325.2
Community Design Council~1K power users285%11.7
06 — PRD Excerpt

Product Requirements

Problem

Figma's UI3 migration and default opt-in AI training created significant friction with power users. Trustpilot rating in the 'Poor' range (~2–3★), dozens of forum threads documented widespread backlash, and a proposed class-action lawsuit was filed over AI data practices. Power users report noticeable productivity drops and feel their feedback is not being heard.

Goal

Restore Trustpilot rating to 3.5★+ within 180 days. Reduce UI-related support tickets by 35%. Achieve 90%+ opt-in to new features within 120 days of launch (voluntary, not forced). No additional AI-related legal actions.

Users

Millions of MAUs across tiers. Primary: power designers (heavy daily users on paid Professional/Organization seats). Secondary: design-adjacent roles (PMs, devs, marketers) who represent ~2/3 of user base per S-1. Tertiary: enterprise admins concerned about IP protection.

Metrics

P0: Voluntary feature adoption rate ≥90% within 120 days. P1: Forum sentiment shift from majority-negative to ≤20% negative threads. P2: AI consent dashboard engagement ≥50% of paid users. Guardrail: No increase in churn among $100K+ ARR accounts.

Non-Goals

Not reverting UI3 entirely. Not removing AI features. Not matching Penpot's open-source model. Not eliminating Dev Mode as a paid feature.

Risks

Configurable workspace increases maintenance surface. AI consent dashboard may reduce training data volume. Gradual adoption may slow feature velocity. Community Council could amplify rather than resolve complaints.

07 — Impact

Projected Outcomes

Trustpilot Rating
~2–3★
3.5★+
Exit 'Poor'
Voluntary Adoption
0% (forced)
≥90%
Agency restored
Forum Sentiment
Majority negative
≤20% negative
Trust rebuilt
AI Consent Engagement
~0% (hidden)
≥50%
Transparency

180-Day Validation Plan

Day 0–30
Workspace Controls
Ship toolbar position + panel docking toggles. A/B test on 10% of Pro users. Measure task completion time.
Day 30–60
AI Consent Dashboard
Account-level opt-in/out with data usage log. Track engagement rate and opt-out percentage.
Day 60–120
Gradual Feature Intro
Pilot opt-in toggle for next major feature. Measure voluntary adoption curve vs. forced baseline.
Day 120–180
Full Rollout + Council
All workspace controls GA. Design Council launches. Monitor Trustpilot, forum sentiment, NPS.
What I Learned

The UI3 backlash was not only about the redesign itself, but about how quickly choice was removed. The AI controversy was not simply about using AI, but about how customer work was included in training with defaults many users did not expect. Both moments highlight the same underlying risk: it is possible to prioritize shipping velocity in ways that leave power users feeling like they have less agency than before.

Growth metrics can temporarily hide early signs of trust erosion.
As a PM, I try to design rollouts so that the people
who helped build the moat always feel heard.

Case study by Chetan Jonnalagadda · chetanjonnalagadda.com

Financial data from Figma FY2025 earnings. Quotes are representative composites. RICE scores, projected metrics, and solutions are illustrative PM exercises.