I Cloned My Boss's Design Reviews (And She Actually Loved It)

From 3-day delays to instant feedback using AI

Left side shows thousands of collected CTO comments from Figma, right side shows the AI bot interface that learned to replicate her feedback style and tone

Transforming thousands of CTO comments into an AI that speaks and critiques just like her

Project Overview

Project: Eliminated CTO review delays by systematically capturing institutional design knowledge

My Role: Process efficiency designer identifying and solving workflow constraints

Timeline: 1 month (May 2024) - from “this sounds interesting” to “why didn't we do this sooner”

Impact: Cut down 2-3 day review delays per design cycle, saved 3 hours/week per designer

Design Challenge & Approach

The Challenge

Our CTO gave excellent design feedback, but her schedule created 2-3 day delays for every iteration. Review scheduling was becoming a team velocity constraint.

My Solution Strategy

I thought: “What if you could get expert design feedback instantly instead of waiting for review meetings?” I built an AI clone of our CTO's review style that eliminated design bottlenecks while maintaining quality standards.

Technical Process & Solutions

Part 1: Knowledge Extraction

Problem: Years of design expertise lived only in scattered Figma comments across 300+ files

What I did:

  • Used AI to figure out API extraction scripts to gather all the comments by the CTO and also by anyone else who mentions “the CTO said..”
  • Used batch processing to handle the scale effectively
  • Successfully scraped over 10,800 comments by the CTO or referring to the CTO's feedback
Animated demonstration showing the extraction of design feedback comments from Figma files, displaying the process of gathering institutional knowledge from scattered comments across multiple design files

Extracting years of design expertise from scattered Figma comments across 300+ files

Part 2: Creating “Design Review Assistant”

Problem: Generic AI feedback lacks the nuanced understanding that makes expert review valuable

What I did:

  • Integrated the CTO's documented work style preferences and Storehub's design principles and user demographic information to provide further context
  • Structured around her proven review framework covering technical, user experience, and business considerations
  • Maintained the collaborative tone that makes feedback constructive rather than critical

Plot Twist: When Your AI Gets Too Realistic 🤖

One unexpected challenge emerged when team members tried using the bot on Fridays. Because I'd included the CTO's user manual (which specifies no-meeting Fridays for deep work), the AI would sometimes respond with “Hey I am actually doing deep work so this better be important.”

While amusing, this taught us an important lesson about context filtering: sometimes you capture TOO much authenticity. We had to fine-tune which personality traits enhanced the tool versus which ones created unnecessary friction.

Screenshot showing the AI design review assistant in action, demonstrating how it captured the CTO's authentic feedback style including personality quirks that sometimes created unexpected responses

The AI clone captured authentic feedback patterns - sometimes a little too authentically

Results & Organizational Impact

AI Design Review Assistant final implementation showing the comprehensive system that successfully eliminated review delays while maintaining expert-level feedback quality

The final AI design review system that transformed our team's workflow and feedback process

Measurable Efficiency Gains

This design review assistant is able to provide early context and feedback. The 50% reduction in direct CTO comments represents significant time savings for strategic work, plus elimination of iteration delays that were slowing team progress. This bot can also help with scaling work - whether we're 3 designers or 30, quality standards remain consistent without scheduling constraints.

What This Actually Accomplished

Beyond the novelty, this project solved a real workflow challenge. By reducing design iteration cycles from 5 days to 2 days, we were able to enable 40% faster feature delivery during critical compliance deadlines. It's also a great bonus that the CTO loved it, because the bot actually understood her review philosophy.