Pulse
A Pulse Field Note · Compass

The AI feature your power users will quietly resent

We showed the same AI edit to pros, hobbyists, and social creators — and the people who should love it most pushed back hardest.

Panel
Photographers · 80 moderated sessions
Protocol
9 workflows · moderated UR + multi-turn
Records
160 sessions · 488 workaround citations
Field date
May 1, 2026
The anatomy of this study
Test Cases
×
Persona Agents
×
Metrics
×
Business Context
=
Contextual
Evaluation
01 · Test Cases

Realistic scenarios & goals

9 real editing workflows · moderated + multi-turn
The goalSurface the friction and workarounds a launch can't see in advance — where a new capability reads as a gift, and where it reads as a threat to craft.
02 · Persona Agents

Who we put on the panel

stateful, expertise-mixed

Photographer personas spanning professional, hobbyist, and social-creator expertise, run as moderated sessions so the friction is observed, not guessed.

160
sessions
2
image models
9
workflows
03 · Metrics

What we measured, & why

friction · workarounds · silent failures

Counts the moments users hesitate, redo, or route around the feature — the signals that predict churn but never show up in an aesthetic score.

How they scored

Friction events
477
Workaround citations
488
04 · Business Context

Why this matters

what makes the finding meaningful

Beta channels and in-app feedback skew toward power users, so the middle of the market gets missed before launch. This study surfaces the segment split a launch can't see in advance.

Knowing which segment a feature delights and which it quietly alienates is the difference between a celebrated launch and a feature your best users learn to avoid.

Summary
Verdict
The same generative edit reads as a gift to some users and a threat to craft for others. Across real photographer workflows it drew 477 friction events and 488 workaround citations — most from the users with the highest craft expectations.
Method
Two image models (GPT Image 2 vs Gemini 3 Pro Image) run through real photographer workflows — deadline retouch, client revision, series consistency, shot recovery, mood transform — as moderated user-research sessions plus a multi-turn iterative-edit study. 160 sessions across 9 workflows.
Knockout
488 workaround citations — the moments users quietly route around the feature rather than trust it — across just 9 workflows.
Surprise
The friction isn't broken output; it's the honest-but-exhausting moments where the tool makes a pro redo work they'd already nailed.
Limitations
Two models, photographer-centric workflows, mixed moderated and simulated sessions. Friction counts are directional signals of where to look, not a defect rate.
Go deeper — the full study, every persona, and all the evidence.
See the full study →