Vignesh G. Let's Talk →
← Back to stories AI Design Review · Self-Initiated

Method's AI design-review copilot

Method's design team works across time zones - a designer often waits on a peer's feedback before validating work with the client, or wants an independent check before that client review even happens. There was no consistent way to check a design against every angle in between. So I built a copilot that does exactly that.

Company
Method / GlobalLogic
Role
Design Lead, tool built for internal use
Tool
Method UX Review Copilot
The Opportunity

Design validation at Method depended on who was awake. A designer in one region would wait on a peer in another to weigh in before taking work to the client; other times, they wanted an independent check before that client review even happened. Either way, there was no consistent way to validate a design against every angle - usability, accessibility, trust, business impact - without waiting on someone else's calendar.

The Initiative

Design validation was informal and person-dependent - you got as thorough a review as whoever had time to give one.

I built a small AI-assisted copilot for the global design team to validate work against six lenses on demand, so a designer could check every angle before taking it to the client instead of waiting on a colleague first. It wasn't part of my workstream brief; I ran it alongside my actual responsibilities because the gap was clear.

Leadership Scope
Cross-functional collaboration
Self-initiated the tool outside my formal workstream scope, because the validation gap was clear
Built it solo, then shaped the review governance model - it flags, a person still decides
Drove adoption across the global design team, spanning three time zones
Critical Product Decisions
01
Kept a human in the loop, not a gate
The team almost treated it as a pass/fail gate. I pushed back - it flags, a person still decides. AI didn't change what good design is, it changed how easily a designer could validate their own work before the client saw it.
02
Reviewed the whole journey, not screen by screen
Up to 20 screens go in together, grounded as one continuous flow instead of independent findings per screen - closer to how a real design review actually reads a product.
03
Rolled every finding up into business risk
Every finding closes with a business-impact tag - conversion, retention, support cost - so the review lands as a business conversation, not just a design one.
How It Works
01
Configure
Describe the project brief, pick the standards that apply - WCAG 2.2, EAA, ADA, Nielsen heuristics, Material, HIG - or apply a saved preset.
02
Upload
Drop in up to 20 screens as PNG or JPG. The engine grounds every finding in the actual interface - reviewed together as one journey, not screen by screen.
03
Receive
A board-ready report: one overall score, six lens scores, visual callouts, prioritized findings, and a roadmap grouped by effort and impact.
Evidence
The Tool
overview · six lenses, one verdict
Method UX Review Copilot overview page showing the six review lenses
Usability, accessibility, trust & ethics, business impact, design quality, and Method's own experience framework - combined into one report instead of five separate ones.
configure · pick your lenses
New review configuration screen showing selectable review standards across UX, accessibility, trust and ethics, and design systems
Every standard is selectable - WCAG 2.2 AA, ADA, EAA, Nielsen heuristics, dark-pattern detection, even a specific design system. This run: 12 standards, Full review depth.
multi-screen · the whole flow
Eight J-ESTA screens attached together for one unified journey review
All 8 J-ESTA screens attached and reviewed as one continuous journey - home through terms & conditions - not eight disconnected reviews.
The Result - Six Lenses, One Verdict
the verdict · 68/100
Overall UX score of 68 with six lens scores and an executive summary
One score, six lenses, needs-attention flagged immediately - Accessibility (58) called out as the weakest lens before you read a single finding.
usability lens · a finding
Critical usability finding: placeholder text used as form labels, with issue, recommendation, and business impact
Every finding is severity-graded, tied to a real screen location, and paired with a recommendation and a business-impact tag - not just "this is wrong."
design quality lens · strengths
Four design strengths worth preserving, each with why it works and its business benefit
Every review also surfaces what's working. A second opinion, not just a fault-finder - the strengths get a business case too.
accessibility lens · WCAG 2.2 AA
Accessibility review mapped to WCAG 2.2 AA, ADA, and EAA standards, broken down by contrast and touch target findings
Mapped to real standards - WCAG 2.2 AA, ADA, EAA - and broken down by category: contrast, touch targets, keyboard, screen reader. Not a generic accessibility score.
trust & ethics lens
Trust and ethics review showing transparency, user control, and consent scores, plus dark pattern risk
Dark-pattern risk, consent clarity, and user control - the lens most review tools skip entirely.
method experience lens · proprietary
Method Experience review, a proprietary six-dimension lens covering value clarity, trust building, guidance, action clarity, friction, and business alignment
The one lens that's ours: does the screen build trust, guide the user, and align to the business goal - Method's own 6-dimension framework, not a borrowed heuristic.
business impact lens · the roll-up
Business impact summary showing risk areas across conversion, adoption, retention, trust, support costs, and task completion
Every finding rolled up into business risk - conversion, retention, support cost - so the review closes as a business conversation, not just a design one.
Outcome

The global design team adopted it as a standard step before taking work to the client, not a mandate I pushed through - it earned that by giving designers a way to validate on their own timeline instead of waiting on a colleague first.

Reflection
Easy to overclaim a tool like this. What actually happened: a handful of designers across regions started validating their own work before it went to the client, instead of waiting on someone else's feedback first. That's a real behavior change, even a small one. I'd rather be precise about that than inflate it.
Next story
Designing a digital ID that replaces the passport in your pocket - JESTA

Curious about where AI actually helps vs. where it doesn't - happy to talk through it.

Let's Talk → Back to all stories Resume LinkedIn