Blog · Case studies
When a retailer gets tens of thousands of Messenger conversations a month, answering faster is not enough. You have to know which conversations are real buyers. Here is how that system was approached.
By Dijae Dadula · October 13, 2026 · 4 min read
Most chatbot case studies are about small businesses getting a few more replies a day. This one is about scale. EasyPC is a large Philippine computer retail operation, and its Messenger inbox looks nothing like a typical small business inbox.
At the time of this work, EasyPC's operating context looked like this:
16 branches across its retail network.
10,000+ SKUs, from components to full builds and accessories.
About 1.3 million followers on its social pages.
About 40,000+ monthly Messenger contacts.
That last number is the one that matters. Forty thousand conversations a month is more than a thousand a day. Many of them are serious buyers. Many are not. And from the inbox, they all look the same.
The obvious answer to a busy inbox is "reply faster." But at this volume, speed alone just moves the bottleneck. If every message gets the same fast reply, the sales team still has to read thousands of conversations to find the people who are ready to buy.
The real challenges were:
Identifying real buying intent among price checks, stock questions, support issues and casual browsing.
Separating meaningful leads from noise so staff time went to the right conversations.
Structuring lead information so a qualified lead arrived with what the team needed to act: what product, which branch, how soon.
Keeping answers consistent across a huge catalog and many staff members.
Routing conversations to the right place instead of one shared pile.
This is the pattern I describe in Don't Automate Chaos: The 8-Step Method I Use Before Building Any AI System: start with the business problem and the current process, not the tool.
Instead of treating the chatbot as a fancy auto-reply, the work treated every conversation as something to classify and qualify. The building blocks were:
Structured qualification. A defined idea of what a "perfect lead" looks like for the business, and questions that move a conversation toward that picture without feeling like a form.
Classification. Conversations sorted by what they actually are: buying intent, product inquiry, order or support issue, or noise. Each type gets a different path.
Automation for the repetitive layer. The questions asked thousands of times a month get answered instantly and consistently.
AI-assisted responses for the messy middle, where customers ask things in their own words and a button flow would break.
Structured lead records. Qualified conversations end up as organized lead information, not just a chat transcript, so the team can act on them quickly.
The tools in this kind of Messenger-first setup are usually Pancake for the shared inbox and Botcake for chat automation (affiliate links). I compare them in Pancake vs Botcake vs ManyChat: Which One Fits Your Business?.
The project has been associated with reported improvements of approximately 300× in qualified leads, 500× in lead quality and 9× in response speed. These are the project's reported figures, measured in its own context against how the inbox was handled before. They are not a promise of what any other business will see, and the baseline matters: when a process starts with almost no structured qualification, the improvement multiple can be very large.
What is more useful than the multiples is the principle behind them. The gains did not come from "adding AI." They came from deciding what a qualified lead is, designing the conversation to find it, and connecting that to how the team works.
Define a qualified lead before you automate anything. If your team cannot describe a good lead in two sentences, no chatbot can find one.
Classify before you reply. A price question, a complaint and a ready buyer need different paths.
Measure outcomes, not conversations. More chats is not success. Qualified leads, sales and response time are. I explain how to test and measure this in How to Test an AI Chatbot Before It Talks to Your Customers.
Keep humans where they matter. Automation handles volume. People handle judgment, negotiation and problems.
Build for your scale. A system for 40,000 conversations a month looks different from one for 400, but the method is the same.
It is using automated conversations and AI to ask the right questions, understand the answers, and decide whether someone is a likely buyer, so your team spends its time on the people most ready to purchase.
Yes, if it is designed for that volume: clear classification, consistent answers for common questions, structured lead capture and a clean handoff to people. The design matters more than the tool.
Not automatically. The reported figures come from one business and its starting point. The method, defining and qualifying leads before automating, applies to most businesses that sell through chat.
Messenger-first businesses in Southeast Asia often use Pancake for the shared inbox and Botcake for automation. ManyChat is another strong option, especially for Instagram.
Pancake coupon: use code CHAMP20 for 20% off all Pancake plans plus 30 days free. Valid for new Pancake customers or a new billing account on your Pancake account, and applied at checkout when you subscribe or renew. Get Pancake here
Join the free newsletter on AI agents, automation and turning your skills into income. Prefer video? Watch the walkthroughs on YouTube.
Want a system like this built for your business? Book a free AI Automation Blueprint call, or see the done-for-you services at my consultancy, IRD Digital Marketing Services.
DIJAE DADULA
Dijae Dadula · AI automation and systems specialist · Cavite, Philippines
Home · Blog · Book a call · Newsletter · YouTube · Threads
Some links on this site are affiliate links. If you buy through them, I may earn a commission at no extra cost to you.
© 2026 Dijae Dadula. All rights reserved.
Thank you
Your submission has been received! Keep an eye on your phone or email because we will contact you soon.
