1
Floating helper collects what you need
On Taobao, Tmall, and 1688: tap → pick images → send to Studio.
Product design case study · Live team tool
EvikAI Studio
A reliable workflow that turns a Chinese product page into an Etsy-ready listing — from capture and AI-assisted craft to team review and publish.
2
AI copy + batch image edits
Generate English title, description, and tags; batch-restyle selected images in one place.
3
One click to Etsy — draft or live
Push images, copy, and price. Choose review, draft, or publish.
4
Schedule publish + social promo
Plan go-live on Calendar; share the same listing to social in one click.
01 · CaptureCollect on the source page
02 · CraftCreate copy and images together
03 · PublishReview, schedule, or push to Etsy
01 · Context & problem
The work was fragmented before publishing even began.
Sellers moved between source sites, folders, photo tools, ChatGPT, and Etsy. Every handoff introduced delay — and every mismatch weakened trust.
25–30
Minutes spent collecting a product before copy work started
Images were downloaded one by one, links copied manually, and folders reorganized before the listing could take shape.
Images and English copy lived in separate tools
Batch work became repetitive and context was lost between photo apps and AI tools.
What Studio showed did not always match Etsy
Price, gallery order, variants, and duplicate drafts made every Push feel risky.
One shared login could not support different responsibilities
Creators needed to prepare drafts; publishers needed shop access and a controlled network context.
02 · Research & signals
Test the whole job, not isolated screens.
Exploratory interviews and workflow walkthroughs focused on where users lost time, confidence, or control from product capture to Etsy draft.
01
Batching reduced tool switching
One teammate reported product-image preparation moving from about 30 minutes to about 6 minutes when selection and restyling stayed together.
02
Role clarity reduced accidental publish anxiety
Creators felt safer when their workspace simply did not expose shop connection or Push.
03
A first-run user needed less infrastructure, not more instruction
Replacing the local server and large installer with a small extension made onboarding understandable.
03 · Project goals
Two promises the product had to keep.
The case was deliberately narrowed from a list of features to two product-level outcomes.
Build a seamless path to Etsy publish
The title, images, price, variants, and listing state reviewed in Studio should be what Etsy actually stores.
Let teammates collaborate without sharing risky access
Creators should prepare high-quality drafts while publishers retain deliberate control over shop connection and Push.
1Design challenge
Build a seamless path to Etsy publish.
Tool switching fragmented one listing into many separate tasks
Sellers downloaded images manually, copied source information, moved between ChatGPT, Photoshop or Canva, and Etsy, then entered the same listing data again at publish. Each handoff consumed time and created another opportunity for content to be lost or become inconsistent.
Carry one listing through the whole journey
Bring source content, AI-generated copy, edited images, review state, and publish actions into one continuous workspace — from capture to Etsy — so the product moves forward without manual rebuilding.
How I turned a fragmented workflow into one listing journey
The work moved through six connected steps: understand the real user task, frame the product around that task, redesign the journey, make interaction decisions, test the whole job, and separate evidence from assumptions.
01
User problem · map the real workflow
I reconstructed the current journey from product discovery to Etsy publishing, marking every download, copy-and-paste, asset handoff, repeated field, and point where the user had to remember what was already done.
Source sites→
Downloads→
ChatGPT→
PS / Canva→
Folders→
Etsy
What the map revealed
- Context reset at every tool switch; users repeatedly rebuilt product meaning.
- Images, copy, and product data had no single owner, creating version uncertainty.
- There was no visible definition of “ready,” so progress lived in the user’s memory.
02
Product definition · reframe the job
The interviews were not really describing an image-editing problem. Users were trying to finish one commercial object — a publishable listing — while the product forced them to think in tools and files.
How might we let sellers think about one listing instead of managing five separate tools?
Product principle
Make the Listing the center of the product. AI copy, image preparation, review, and publishing become capabilities attached to it. Source context and decisions travel forward with the same cloud draft instead of being rebuilt at each stage.
03
Journey redesign · simplify the mental model
I replaced the tool-based sequence with three user-readable stages. The interface can change by stage, but the underlying Listing remains persistent so users always know what object they are advancing.
CaptureWhat source content do I need?
CraftIs the listing complete and reviewable?
PublishWhat should happen next?
Continuity rules
- One Listing ID remains the source of truth across the journey.
- Saved state makes progress visible and resumable instead of memory-dependent.
- Every next action begins from the reviewed draft, not a blank Etsy form.
04
Interaction design · four decisions
The journey model became four concrete interface decisions, each removing one handoff rather than adding another dashboard or instruction.
1 · Start at discoveryA floating helper captures selected content without manual downloads.
2 · Listing as containerSource, copy, images, price, variants, and status stay together.
3 · Batch repeated workSelect many images, apply one prompt, review the set together.
4 · Choose an outcomeSave draft, publish now, fill in Etsy, or schedule publish.
Why these decisions matter
Together, they reduce context rebuilding at the start, version drift during creation, and uncertainty at the end. The user moves one Listing forward while the system carries the operational complexity in the background.
05
Whole-task validation · test behavior, not screens
The task was intentionally end-to-end: independently move one real product from a source page to an Etsy-ready draft. This exposed whether the redesigned mental model survived the transitions between capture, creation, review, and publishing.
What I observed
Could they find the starting point?
Did they download files manually?
Could they tell what was complete?
Did they leave Studio to finish work?
Did publish choices match intent?
Could they resume the same draft?
06
Evidence · directional internal signals
The pilot did not claim broad market validation. It provided operational evidence that the redesigned path was learnable and removed measurable work inside the team’s real workflow.
~30 → ~6minutes for one product’s image preparation
First runcapture and save completed without deployment help
01
02
03
04
2Design challenge
Separate making from publishing — without breaking the workflow.
Shared publishing access made creators afraid to act
When every teammate saw the same shop connection and Push controls, creators could not clearly distinguish safe preparation from irreversible publishing. Shop owners also worried about exposed credentials and accidental go-live actions.
Separate permissions in the interface, not in a hidden policy
The orange Client workspace supports capture, creation, and saving only. The purple Studio workspace owns shop connection, review, and Push. Both roles collaborate through the same shared cloud draft without sharing risky access.
Production workspace
For teammates who collect products, create English content, and prepare product images.
- Extension capture and image upload
- AI copy and batch image work
- Save to shared cloud history
- No shop connection or Push
Publishing console
For teammates responsible for shop access, listing completion, and deliberate publishing.
- Sync shared drafts from the cloud
- Complete price, category, and shipping
- Connect the Etsy shop
- Review, save draft, or publish
UX information architecture
Role-based entry points keep creation, publishing, and administration understandable without exposing every capability to every user.
EvikAI Studio
Client portal
Studio (edit)
Listings
No Push · No Etsy API
Studio portal
Studio
Listings
Calendar
Help
Admin
Users · Quotas
Studio workspace hierarchy
Capture in
Listing fields
Images
Variants
Push to Etsy
Save / Schedule
Shop settings
Priority: the primary action, Push, sits above editing chrome. Client omits Push and API connection to protect shop credentials.
DDevelopment challenge
Turn the workflow model into a secure, scalable system.
The infrastructure was harder to use than the product
The first version required one computer to remain online and teammates to use remote desktop. The installer that followed removed that dependency, but it was large, slow to set up, difficult to update, and vulnerable to machine-specific bugs — keeping nontechnical users from starting confidently.
Move the product to the cloud and keep the browser lightweight
A small Chrome extension handles source capture while the website, shared drafts, credentials, and API connections live in the cloud. Teammates only need an account and the extension; sensitive keys remain on the server, and everyone receives updates at the same time.
From “keep your own computer running” to “install one small extension”
The architecture changed because the onboarding experience kept failing. I treated infrastructure as part of the user journey: every deployment model changed who could start, how independently they could work, and how safely the team could collaborate.
01
Observe the access problem
The first product lived on one local computer. Teammates could only work by remotely controlling that host, so access depended on one person’s device, schedule, connection, and willingness to keep the machine running.
User impact
- Nobody could start independently.
- Remote control added delay, awkwardness, and interruption risk.
- Sleep, restart, or disconnection stopped the whole team.
- A new teammate could not complete a first run alone.
02
Define the experience goal
The goal was not simply to move code to a server. It was to make access independent from a specific person, device, or installation environment.
How might we let every teammate start independently, without maintaining a computer, configuring a server, or learning deployment?
Experience criteria
Time to first valueHow quickly can someone begin the first listing?
IndependenceCan they work without waiting for another teammate?
RecoverabilityCan work continue after sleep, disconnection, or updates?
03
Iteration 1 · ship a local installer
A downloadable installer moved the program onto each teammate’s computer. This removed the shared remote session and allowed people to run the product without keeping the creator’s machine online.
What improved
Device ownership became distributed: every teammate could install and launch their own copy.
The new bottleneck
Independence improved, but onboarding became harder. The package was large, installation was slow, machine differences caused bugs, and every update required another download. Nontechnical teammates still hesitated to begin.
Learning
Removing one dependency had created a different one: every user now had to maintain a local environment.
04
Final model · Extension + Cloud
Instead of adding more installation guidance, I removed infrastructure from the user journey. A lightweight Chrome extension handles capture; editing, shared drafts, credentials, API connections, and updates live in the cloud product.
Users needed less infrastructure, not more instruction.
Source page→
Chrome Extension→
Cloud Listing→
Studio→
Etsy
Responsibility split
Chrome ExtensionRead the source page, collect selected content, and send it to the user’s cloud draft.
Cloud productStore listings, run AI and image tools, protect keys, connect APIs, and deliver updates to everyone.
05
Validate the first run
The test covered the complete onboarding task: install the extension, capture one real product, open Studio, and save the first Listing without help. This tested the experience rather than merely confirming that the system deployed.
Observation and evidence
Could they identify where to start?
Did they configure a server?
Did they need remote control?
Was developer help required?
A first-run teammate installed only the extension and completed capture plus save without deployment help. The team no longer depended on a 24/7 host computer, remote desktop, or per-device product updates.
Technical architecture
The browser stays lightweight while authentication, credentials, and external API calls remain on the server.
Users
- Browser (website)
- Chrome extension
- Client · Studio · Admin
VPS · EVIKAI.COM
- Node.js server
- Port 443 · website (HTTPS)
- Port 3001 · app API
Connected APIs
- Etsy Open API
- OpenAI · Gemini
- Stripe · ZPAY / WeChat
- Pinterest · Instagram
What each connection does
- Etsy API — connect the shop, Push listings, and read sale dates.
- OpenAI / Gemini — generate English copy and restyle images.
- Stripe · ZPAY / WeChat — manage credits and plans.
- Pinterest / Instagram — create one-click promotional posts.
Extension → :3001
- Capture from Taobao / 1688
- Server talks to external APIs
- The browser never holds secret keys
Etsy OAuth and API access are enabled only for Studio accounts; Client cannot automatically inherit the shop connection.
Why the architecture changed
The first version depended on a local Chrome session and a computer acting as a server. The next version added a large installer and rented VPS. The final model reduced onboarding to a small extension plus the cloud product — a design decision that made collaboration possible, not a technical showcase.
04 · Validation & iteration
Real teammates, real feedback, iteration.
Summarized from internal pilots; names are simplified. Each test asked whether someone could finish capture → craft → Push independently.
yiyi
Studio lead · Etsy shop operations
“Shop connect once landed on a broken address — I thought the product was dead. After the fix, I chose ‘save draft only,’ and our Christmas décor listing finally reached the shop.”
Outcome: shop authorized + first draft Push succeeded
Lily
Content teammate · Client
“Background removal used to take 30 minutes per product. Now I select six images, Restyle once, save — and publishing sees the result immediately.”
Outcome: ~30 min → ~6 min per product’s images
Yang
Partner · Publish + admin
“I worried assistants would publish by mistake. The orange side has no publish button — that helps. Admin also shows everyone’s AI usage.”
Outcome: role split validated + quotas visible
Mina
New hire · First solo run
“I only installed the extension. On Taobao I tapped the floating helper, Studio opened, I followed the purple hints — and saved successfully on the first try.”
Outcome: first capture + save completed without deployment help
How feedback changed the product
Heard
“I only pushed once — why are there two Listings?”
Changed
Only intentional Save or Push creates a Listings row.
Result
The team can identify the real draft without cleanup.
Heard
“I don’t know when my shop sale ends.”
Changed
Show sale start and end dates directly on Calendar.
Result
The next promotion can be planned without digging through Etsy admin.
Heard
“I’m worried an assistant could publish by mistake.”
Changed
Remove Push entirely from Client; keep shop access in Studio.
Result
The role boundary is visible in the interface, not hidden in policy.
05 · Outcomes
From a demo on one machine to a daily team workflow.
The strongest outcome was not the number of features. It was that teammates could move one product through the system and trust the Etsy draft at the end.
The product aligned three truths: seller intent, Studio state, and Etsy output.
The current evidence comes from internal pilots and repaired live drafts. It supports product direction, while broader seller research remains the next validation step.
Directional pilot evidenceMinutes for product-image preparation
Reported by one content teammate after batch selection and restyling were combined.
Small extension instead of a local server setup
A first-run teammate completed capture and save without deployment knowledge.
Clear workspaces for making and publishing
Client prepares content; Studio controls shop connection and Push.
06 · Reflection
What I learned — and what I would test next.
The 0→1 journey made platform constraints, team roles, and product trust part of the same design problem.
What I learned
- Trust beats polish when an external platform owns the final state.
- Platform rules such as image rank and inventory behavior are part of UX.
- Permissions become easier to understand when the product model makes them visible.
What can be better
- Run structured usability tests with external Etsy sellers, not only internal teammates.
- Instrument capture time, publish success, and correction rate across the full funnel.
- Prove reliable Push before expanding secondary creation features.