I Was The Bottleneck, So I Prompted My Way Out

Picture this: You're the sole designer supporting two product development squads building a credit insurance analytics portal. Deadlines are tight. And then someone asks you to change "Filters" to "Advanced Filters" across 30 screens. 

Again. 

I'd become a bottleneck. BAs, Team Leads and POs would leave Figma comments or message me asking for copy updates because the business terminology was incorrect or they simply wanted to see how copy changes would look. Reasonable request, unreasonable drain on my time. 

Surely Someone's Solved This Already? 

I looked at existing Figma plugins. Most used Google Sheets, which seemed perfect until I hit the security wall – they required sheets to be "viewable by anyone with the link." Great for your personal side project. Terrible for sensitive company data that absolutely should not be floating around the internet. 

Other solutions were built for bulk data replacement – think populating table rows or product catalogues. Their spreadsheet structures were completely unmanageable for what I needed: simple, single-instance text replacements across varied screens. 

What I actually needed was super simple: Column A = label names, Column B = new values. Excel files that could live safely in the company SharePoint. A plugin that could find text layers named #labelname and update them. That's it. 

So I decided to build it myself. Small problem: I'm a designer, not a developer. 

Enter Claude 

Here's where this could have gone very wrong. I could have spent weeks watching YouTube tutorials about TypeScript and the Figma Plugin API. I could have given up entirely. Instead, I tried something different: I used Claude as a collaborative partner, and by that I mean genuinely collaborative: 

Me: "I need a Figma plugin that reads Excel files and replaces text in layers named with a # prefix. Here are the specific conditions..." 

Claude: Generates initial code 

Me: "I'm getting TypeScript errors about Promise and figma not being defined." 

Claude: "You need the Figma plugin type definitions. Here's what's missing..." 

Me: "The UI locks up for 10 seconds before showing progress. Can we add a loading state?" 

Claude: "That's the document scanning phase. Here's how to make it yield control and show a spinner..." 

This continued through bugs, features, refinements. I brought the UX thinking and requirements. Claude removed the barrier of implementation. I stayed in control of what and why. Claude handled the how. 

The Result: SheetSync 

Within a few hours, I had something working. A bit of refinement later, and the team now maintains a shared Excel spreadsheet with all screen copy. When they want to explore changes, they update the sheet, and I drag it into SheetSync. The plugin updates everything in seconds. 

What used to take me hours now takes them 30 seconds. I get my time back for design decisions. They get immediate feedback. Nobody's waiting on me to change "Login" to "Sign In" for the fourth time. 

The Actual Point Here 

We're living through a period where AI is being crammed into everything whether it makes sense or not. Your toothbrush probably has AI now. Your fridge is definitely "AI-powered." Most of it is marketing nonsense. 

But here's what I learned: when you have a real problem and you use AI as a tool (not a replacement for thinking), it can actually be transformative. Even if you're not technical. Especially if you're not technical. 

I'm not a developer, but I could clearly articulate what I needed and why. I could test whether solutions worked. I could iterate on details. The AI removed the coding barrier, but I still needed to do the actual thinking. 

So no, AI didn't replace my job. It just gave me back the time to actually do it. 

SheetSync is now used across both product squads. I've reclaimed valuable time that I'm now spending on actual design work instead of being a human autocorrect function. 

If you're facing similar bottlenecks in your workflow, maybe it's time to stop asking "should I learn to code?" and start asking "what could I build if the coding part wasn't the barrier?"