CASE STUDY
Finding 200 buyers in 30,000 DMs - without a chatbot
A reel goes viral. 30,000 people flood the DMs of a premium metal-card brand. Most are tourists. Buried inside that flood are maybe 200 people who will actually pay - and a team of three trying to find them by hand.
4x
Conversion increase
30k
DMs processed per spike
~200
Real buyers identified
5
System layers shipped
THE CLIENT
CarbonCraft
CarbonCraft makes premium metal cards - NFC-enabled, custom-designed. When a reel goes viral on Instagram, their DMs explode overnight. 10,000 to 30,000 messages in a day. Three people trying to respond. Most messages are tourists. Maybe 200 are actual buyers. Finding them meant reading every conversation, one by one.
Everyone in AI would have shipped a chatbot to “engage all 30,000.” We refused. We went inside the business, found the real shape of the problem - the buyers aren't a crowd, they're a needle - and built a system to find the needle.
Industry
Premium Cards & NFC
Challenge
Viral DM overload
Team size
3 people
Scale
30,000 DMs/spike
Approach
Lead intelligence
Result
4x conversions
THE BRAIN
Lead intelligence that reads intent like a veteran salesperson
Not a chatbot - a scoring engine that watches every conversation and understands who's about to buy. Every customer gets a live 0-100 intent score, assembled from four lenses.
Engagement - did they send a photo of the card they want? Ask a specific design question? Reply within 5 minutes? Purchase intent - “how much?”, “do you deliver?”, “is COD available?”, urgency, comparing products. Profile - returning customer, referral, bulk inquiry. Negative signals - price-and-vanish, silence for 24 hours, spam. These subtract.
Scores decay over time if a lead goes quiet. When someone crosses into Hot (80+), an alert fires. Out of 30,000 strangers, the team opens their screen and sees a ranked list of the few who matter, hottest first.
“Everyone else builds a bot to talk to 30,000 people. We built intelligence that finds the 200 who'll actually buy.”
THE COCKPIT
One screen that runs the entire company
The intelligence is useless if the business can't act on it. So the whole company lives on one real-time dashboard.
A live inbox - WhatsApp and Instagram in one place - where every conversation streams in over WebSocket. Open a chat and see who this person is, their live lead score, their order history, their sentiment, and suggested replies.
An order pipeline (Kanban) where every order moves down one rail: Confirmed, Design, Production, Quality Check, Shipped, Delivered. Payments, products, customers, analytics - revenue trends, the AI-versus-human split, the live sales funnel - all in one view.
Plus a team-performance system with leaderboards, response times, SLAs, and targets.
Sales, design, production, shipping, support - five jobs, one screen, moving in real time.
SELF-HEALING INFRASTRUCTURE
Infrastructure that reads its own code and fixes itself at 3am
A dedicated watchdog checks every part of the system every 60 seconds. When something breaks, it doesn't just alert - it reads the codebase with Claude, diagnoses the root cause, repairs or restarts the failing service itself, and texts the founder on WhatsApp: handled.
“At 3am a service stumbled. No human woke up. In 60 seconds the system noticed, read its own code, fixed itself, and sent one message: resolved.”
THE RESULT
From drowning in DMs to converting on autopilot
Before: ~5-6 buyers
After: ~20-30 buyers per viral spike. Same team of three.
Before: 30k DM chaos
After: A ranked list of real buyers, hottest first.
Before: Sales dead between spikes
After: Team works the warm backlog - steady conversions even when nothing is trending.
Before: 5 separate tools
After: One cockpit. Sales, production, shipping, support - live.
THE PRINCIPLE
We don't deploy generic AI. We build the exact system a business actually needs.
We didn't build a chatbot. We didn't “add AI” to their existing workflow. We went deep enough to understand that the problem wasn't engagement - it was identification. The buyers were already there. They just needed to be found.
That's the Veyu thesis in one project: go inside the business, find the real shape of the problem, and build a system whose only job is to solve it.
GET STARTED
Ready to put AI to
work for your business?
Book a free consultation. We'll map out exactly where AI saves you time and money, and you'll walk away with a plan whether you work with us or not.