The London Brow Company sells professional-grade brow products online. It's a beautiful brand with quality products, and it's growing fast.
One question was eating its support team alive: "Which shade is right for me?"
This case study covers how we studied the business, found the real operational bottleneck, and built an AI that cut shade inquiries by around 90%.
The Problem: One Question, Thousands of Times
When we first spoke to The London Brow Company, they called their support inbox "Groundhog Day."
Every day, hundreds of customers asked some version of the same question:
- ●"I have dark brown hair, which shade should I get?"
- ●"I'm between blonde and light brown, what do you recommend?"
- ●"I bought Ash Brown before but it was too warm, what's cooler?"
- ●"Can you help me match my shade? Here's a photo..."
The support team had become full-time shade consultants, and every one of these questions was a buying signal. The people asking weren't complaining or returning anything. They were ready to buy and stuck at the finish line because they couldn't pick a color with confidence.
The numbers were brutal:
- ●400+ shade inquiries per week
- ●Average response time: 6 hours
- ●Support team spending 70% of time on shade questions alone
- ●Estimated abandoned carts from unanswered shade questions: hundreds per month
The usual answer would be to hire more support staff. That only throws labor at a process that shouldn't need people in the first place.
Our Approach: Study the Business, Solve the Real Problem
Custom AI differs from a generic chatbot in where it starts. We start with the operation, and the technology comes later.
Before writing any code, we spent time on The London Brow Company's specific situation.
What we studied:
- ●The product line: Every shade, undertone, and how they actually look on different hair colors
- ●Historical support tickets: 6 months of shade inquiries, categorized by question type
- ●Customer decision patterns: What information actually helped customers choose confidently
- ●Common mistakes: Which shades got returned most often, and why
- ●Expert knowledge: How their best support rep guided customers to the right shade
What we discovered:
Shade matching followed a clear logic:
- ●Hair color → narrows to 3-4 possible shades
- ●Undertone preference (warm/cool/neutral) → narrows to 1-2 shades
- ●Desired intensity (natural/bold) → final recommendation
- ●Previous purchases → personalized refinement
The support team was running this decision tree by hand, 400+ times a week. That made it an operational process that could be systematized and automated, rather than a customer service problem.
The Solution: AI Shade Consultant
We built an AI agent trained on The London Brow Company's shade logic, a specialist that knew their products in depth.
What the AI could do:
- ●Ask the right questions: Hair color, undertone, desired look
- ●Recommend with confidence: "Based on your medium brown hair with warm undertones, I recommend Chocolate #4"
- ●Explain the reasoning: "This shade is warmer than Ash Brown and will complement your natural coloring"
- ●Handle edge cases: "You mentioned your last shade was too dark, so let's go one shade lighter"
- ●Show visual references: Product images, swatches, customer photos (with permission)
- ●Compare options: "Here's how Chocolate #4 compares to Soft Brown #3"
What made it work:
Instead of rigid rules, the AI learned from how their best support rep actually helped customers, including the rep's follow-up questions and explanations.
When a customer said "I want something natural looking but I have grays," the AI knew that called for a different recommendation than "I want full coverage."
The Results: Around 90% Fewer Shade Inquiries
The build took 5 days, and results showed within the first week.
Shade Inquiry Volume
- ●Before: 400+ shade inquiries per week
- ●After: ~40 inquiries per week (edge cases and complex situations)
- ●Reduction: 90%
The AI handled 360+ shade consultations a week on its own, instantly and around the clock.
Support Team Impact
- ●Before: 70% of support time on shade questions
- ●After: 70% of support time on actual support issues
The team could finally focus on order problems, returns and VIP customers, and morale improved once the copy-paste work was gone.
Answers at 9pm, Not the Next Morning
A customer asking about shades at 9pm used to wait until the next morning for a reply, if they waited at all.
Now AVA answers in seconds at any hour. It handles 80+ after-hours chats a month that used to sit in the inbox overnight.
Customer Satisfaction
Customers didn't complain about "talking to a bot." They complimented the experience:
- ●"That was so easy, I was worried I'd have to wait for an email"
- ●"The recommendation was exactly right"
- ●"Way better than the shade finder tools I've used on other sites"
The AI came across as helpful because it was trained on real expertise instead of generic scripts.
The Bigger Lesson: AI Solves Operational Problems
Most brands look at AI customer support as a way to replace people. The bigger win is finding the operational bottlenecks that never needed people in the first place.
The London Brow Company had a product guidance problem that looked like a customer service problem. Its experts' knowledge lived in their heads, and every customer had to wait for access to it. The AI made that knowledge available instantly, to everyone, at any time.
Every eCommerce business has a version of this:
- ●Fashion: Size and fit guidance
- ●Beauty: Shade matching, skin type recommendations
- ●Electronics: Compatibility questions
- ●Home goods: "Will this fit in my space?"
- ●Food/supplements: Dietary and usage questions
Each one is a decision tree that experts run through by instinct. AI can learn that tree and run it thousands of times at once.
How We Approach Every Project
The London Brow Company project follows our usual process:
- ●Study the business: We start with your operation, products and customer behavior, and bring in AI after that.
- ●Identify the bottleneck: What process is eating the most time? What questions repeat endlessly? Where are customers getting stuck?
- ●Map the expert logic: How do your best people solve this problem? What questions do they ask? What information do they need?
- ●Build a specialist AI: An AI trained on your business, products and logic, instead of a generic chatbot.
- ●Measure and refine: Track what the AI handles and what it escalates, and keep improving it.
The result is an AI that works like your best employee, available 24/7.
Have an operational problem disguised as customer support? We'll study your business and build an AI that actually solves it. Book a strategy call