← Back to portfolio

AI agent · Figma plugin now live

UX Doctor Agent

A diagnosis partner for struggling products

I'm building an AI agent that works next to me on every UX audit. It researches the product, tracks the evidence and remembers every past case, so I can find the real problem behind the symptoms in about 48 hours. The diagnosis and the decisions stay mine.

Being built and tested on live audits
Free · Figma CommunityTry the UX Doctor pluginBuild your own emotion trail from any flow in Figma.

How I diagnose

I follow the emotions, screen by screen

First I talk with the product's owner if I can. If not, I walk through the product step by step and note my first feeling on every screen.

The single feelings matter less than how they change. The place where confidence turns into doubt is where the symptom lives. The agent records this trail for every audit, so each finding can be traced back to the moment it was felt.

An emotion trail

Example · a crypto wallet signup
confident lost

Example for illustration, not data from a real product.

Principles

What the agent learns from me

These are the rules I diagnose by, in my own words. The agent applies them on every audit and flags any place where a product breaks one.

01Treat every product like an everyday object▾

"I diagnose any app, website or product as if it were the simplest thing everyone uses every day."

A complex domain is no excuse for a complex experience. Anything that makes a person stop, hesitate or ask for help is a symptom.

AgentStarts every audit with a naive first-time-user pass before any expert analysis.

02Keep the science behind the scenes▾

"Simplifying behind the science is the best."

The logic, data and technology can be as complex as they need to be. The designer's job is to absorb that complexity so the user never feels it.

AgentJudges each solution by how much complexity it takes away from the user.

03The business pain should be the user's pain▾

Before anything else I ask two questions: what was this product made for, and who was it made for? Problems are ranked by how directly they block that purpose for those people.

When the business is solving one problem and the users feel another, that gap is the deepest finding of all.

AgentWon't start an audit until purpose and audience are written down.

04Reasons over numbers▾

"When an app is created to be something big, it needs to solve something big."

Completion rates show what happened. People's reasons show the hook, the real need that brings them back. A big product needs a problem big enough to match.

AgentCollects the "why" from reviews, threads and interviews before it looks at metrics.

05Follow the proven path, improve at the edge▾

If a successful product already solves the same problem well, I don't experiment with your users. For a wallet, I follow 80–90% of MetaMask's path and add one thing that is nicer or simpler.

AgentMaps the category leader's flow before proposing any solution.

06Cure the disease, not the symptom▾

A good UX designer works like a doctor. A new button colour is pain relief; a flow that no longer needs the button is the cure.

AgentTraces each symptom down through UI, flow and structure until it reaches a decision someone made.

The 48-hour audit

From first look to a checked diagnosis

Every audit follows the same six phases. At each gate I review the agent's work before anything moves on, and nothing reaches you until I have approved it.

  1. 0–2 h
    IntakeYour product, your users, your business model and the problem you already suspect.Gate · scope agreed
  2. 2–8 h
    ResearchCompetitors, reviews, complaints and market context, gathered by the agent.
  3. 8–16 h
    InvestigationThe emotion trail through your core flows, with screenshots and evidence for each symptom.Gate · symptoms reviewed
  4. 16–24 h
    DiagnosisSymptoms traced to their root causes and ranked against your product's purpose.Gate · diagnosis approved
  5. 24–36 h
    SolutionsTwo or three options for each root problem, with tradeoffs and a way to test them.
  6. 36–48 h
    PackageA short summary for your team, the full audit and the evidence behind it.Gate · ready to send

How sure is each finding?

Every claim in an audit carries an evidence level, so you always know what is proven and what still needs a test.

E0AssumptionA hunch, with nothing to point at yet.
E1ObservedSeen once in a walkthrough or review.
E2RepeatedThe same complaint across many users.
E3Behaviour dataAnalytics, funnels, support tickets.
E4TestedA usability or A/B test confirms it.

Working together

Start with a free look

Free · 20–30 min

Intro call

You show me the product and share whatever you can. I ask my questions.

Free · within 24–48 h

First findings

A follow-up with about half of the problem list, so you can judge the diagnosis for yourself.

Where it helps most

Products with a lot of science behind them

FintechAI productsData-heavy platformsB2B and B2C service providers

The agent does

  • Research products, markets and user complaints
  • Keep the evidence for every finding
  • Remember every past audit and the lessons from it
  • Draft reports and follow-ups for me to review

I always do

  • Talk with you and your team
  • Make the diagnosis and approve every finding
  • Choose and design the solutions
  • Send every message myself

Is your product feeling sick?

Book a free intro call

Write to me with a link to your product and one line about what worries you.

Anahit Khalatyan — UX Doctor for Struggling Products anahit.work