Skip to content
Sign in
Design Review and Critique with Agents

Make critique sharper and more frequent — without outsourcing taste.

Critique is where design quality is actually decided, and it is the part of the process most teams under-resource. This course covers using agents to make review sharper and more frequent without outsourcing taste: structured critique loops, heuristic evaluations and cognitive walkthroughs run at a scale no team could staff, visual regression evidence, accessibility and content review, and design review wired into every pull request.

Pay once. Yours to keep, with updates.

US$39One-time purchase

Buy courseDesigner TrackEverything bundle

Buying for a team? Team seats

Modules
5 modules
Total time
≈3h 50m
Last updated
2026-10-09

Curriculum

5 modules

Slides are the lesson; study guide and transcript come with each module.

Time commitment

≈3h 50m

About ≈4 hours across the full outline — work at your pace.

Access

US$39

Pay once. Yours to keep, with updates.

What you will build / 01

Artefacts you can take back to the team.

Structured critique loops, heuristic sweeps, regression evidence, accessibility and content passes, and design review on every pull request.

01

Critique loops with named dimensions and human final calls

02

Heuristic and walkthrough runs with evidence per finding

03

Screenshot regression boards that do not rely on memory

04

Accessibility and UX-writing reviews as standard passes

05

PR design review with severity levels and escalation

Pinned cream cards in a 2x2 grid with forest-green thread, cobalt tick, and yellow pencil on warm paper
Curriculum / 02

Module outline

Open the first module you have not finished. The slides are the lesson. The study guide and the transcript are the same module if you would rather read.

01

Critique Loops with Agents

The structure of a useful critique loop: named dimensions instead of vibes, the agent as a tireless first reviewer, and the difference between feedback an agent can act on and feedback that needs a human decision.

02

Heuristic Evaluation and Cognitive Walkthroughs at Scale

Two classic methods most teams skip because of cost, run across an entire product by agents: standard heuristics applied screen by screen, task-based walkthroughs step by step, and the human pass that turns findings into decisions.

03

Visual QA and Regression Evidence

Replacing it looked fine last time I checked with evidence: screenshot baselines, agent-run sweeps across states and breakpoints, and regression reports a designer can act on in minutes rather than re-checking everything by hand.

04

Accessibility and Content Review

Two reviews that are chronically deferred, made routine: accessibility sweeps that go beyond automated contrast checks, and UX-writing review that holds copy to the product's voice with the same rigour as the visuals.

05

Design Review on Every PR

The end state: every change that touches the interface gets a design review automatically — token and component checks, screenshot evidence, severity levels — with humans reviewing what matters instead of everything or nothing.

Inside the materials / 03

Sample lesson & slide previews

Titles and diagrams from early modules — the same materials you open after checkout.

Loop diagram of an agent-assisted critique loop. An artifact and its evidence packet feed an agent critique pass run read-only against named dimensions and the brief's review criteria. The critique produces findings, each with severity, evidence, and a smallest recommended fix. A human triage gate accepts, rejects, or marks findings as judgement calls. The agent runs a revision pass on approved findings only, then a re-critique with fresh evidence, and a human makes the final ship decision. A dashed feedback line shows recurring findings being encoded as named dimensions in the harness.Pipeline diagram of evaluation at scale. On the left, two human-led input cards: the product surface (key flows, screens, states, or user tasks worth walking, with screenshots and a live build) and the shared contract (a heuristics file or a task list with personas and the four walkthrough questions, plus evidence rules and a candidate severity scale). These fan out to four agent-run evaluator cards — one per flow for heuristic evaluation, one per task and persona for walkthroughs — which fan back into a single merge agent that dedupes recurring problems, groups them, ranks them by candidate severity and frequency, and lists what was checked and passed. The merged report passes through a human severity pass, where a design lead opens the evidence, confirms or adjusts severity, and rejects unsupported findings, before the confirmed set feeds a fix backlog, a redesign brief, or an archived baseline.Four-column board. Column one: the agent captures baseline and current screenshots for the same routes, states, and widths — 1440, 768, and 390 pixels — shown as paired thumbnails. Column two: one compare agent per page reports observable differences only, each with page, viewport, and evidence file. Column three: every diff is classified as a regression to fix, an intended change that updates the baseline, or capture noise that fixes the capture rather than the UI. Column four: the human review meeting receives ranked blockers with evidence attached, systemic causes grouped to one likely fix, baseline updates to approve, and judgement calls — noise never reaches the room.Three stacked bands. The top band is automated checks run by tools such as axe-core and Playwright: they catch contrast ratios, missing labels and alt text, invalid ARIA, and unlabelled form fields, but miss whether names make sense, whether focus order follows the task, and anything about copy meaning. The middle band is the agent-run semantic and content review: it catches names that do not describe the control, focus order against task order, error messages with no next step, colour-only signals, and terminology and voice drift, but misses real assistive technology behaviour, naming decisions, and legal wording. The bottom band is human judgement: severity calls and exceptions, naming and voice decisions, testing with real screen readers and users, and release trade-offs. Arrows show findings flowing down from each band as evidence for the next.

Critique Loops with Agents · concept

Why critique under-delivers today

Critique Loops with Agents · concept

Named dimensions: what good means, written down

Critique Loops with Agents · concept

The agent as first reviewer

Critique Loops with Agents · diagram

The critique loop

Critique Loops with Agents · example

Findings with evidence, not opinions

Critique Loops with Agents · comparison

Severity keeps the findings list usable

Audience and outcomes / 04

By the end of this course you can…

Design leads, senior designers, and design-minded engineers responsible for review quality — particularly teams where critique currently depends on whoever has spare time that week.

01

Run critique loops where the agent applies named dimensions and the human makes the calls.

02

Execute heuristic evaluations and cognitive walkthroughs across whole products with evidence per finding.

03

Use screenshot-based regression evidence so visual review stops relying on memory.

04

Run accessibility and UX-writing reviews as standard passes rather than special projects.

05

Wire design review into every pull request with clear severity levels and human escalation.

06

Keep taste and final judgement human while making the mechanical parts of review continuous.

Who it is for

This is for you if…

  • Design leads and senior designers responsible for review quality
  • Design-minded engineers sharing review ownership with design
Who should skip it

This is not for you if…

  • People looking for an automated taste oracle
  • Teams that will not put a human on the final call
Why register now

The practice compounds. Waiting does not.

01

Agent output raises volume; review capacity is what actually decides quality.

02

Critique that depends on spare calendar time fails the week drafts get cheap.

03

Wiring review into PRs turns a special project into a standing gate.

Price and checkout

Pay once. Keep the course, with updates.

Prefer the full designer catalogue, or designer plus leadership? Upsell paths use the same checkout you already trust.

Pay once. Yours to keep, with updates.

US$39One-time purchase

Buy courseDesigner TrackEverything bundle

Buying for a team? Team seats

FAQ

Before you register

Will the agent replace design critique?

No. The agent applies dimensions and gathers evidence. Taste and ship decisions stay human — that is the point of the course.

Is this a subscription?

Pay once. Yours to keep, with updates. There is no renewing membership for new purchases.

What do updates cover?

All edits and new lessons in an owned course, for the life of the course. Bundles include new courses added to that track for 12 months.

What if a course is retired?

Course retirement gives access to its replacement.

Related material / 05

The sources behind this course.

The course teaches the practice; the books and articles carry the depth, sources, and worked runs it draws on.

The Agentic Designer

The operating model for product designers, design leads, and builders who need to understand what changes when agents join design work.

Quality, Systems, and Accountability

The three leadership courses are sold together for US$300. Designer courses and bundles don't include them, except the Everything bundle, which does. This purchase doesn't open the designer courses.

View course

Visual QA With Agents

A long-form workflow for using agents to inspect screenshots, compare implementation against intent, run accessibility checks, and produce fix-ready review notes — traced against a real review of this site's own article page.

Read article

Design Critique Loops With Agents

A practical workflow for turning agent critique from vague opinion into structured findings, evidence, severity, revision passes, and human design decisions.

Read article

Design System Audits With Agents

A deep workflow for finding design-system drift across tokens, components, screenshots, copy, interaction states, and product meaning.

Read article

Four Agentic Design Workflows in Production: What Actually Happened

Four real agentic design workflows traced from their own repositories — an editorial site, a multi-agent book pipeline, a book-to-video pipeline, and the design system underneath the site — with the timelines git can prove, the configs that did the steering, the failures the gates caught, and the patterns that repeat across all four.

Read article