About Sidotech
Sidotech is a Vietnamese online retailer specializing in computer accessories and peripherals — mice, keyboards, headsets, storage devices, and gaming gear — distributed exclusively through Shopee, TikTok, Facebook, and Zalo. With a lean team of 25–30, Sidotech fulfills tens of thousands of orders a month, serving a tech-savvy audience of 18–35-year-olds who expect fast, technically accurate support.
The brand's differentiation lies in catalog depth: 100+ specialized SKUs, each with unique driver profiles, firmware versions, and device-specific troubleshooting paths. That same depth made post-purchase support a bottleneck — customers needed detailed, model-specific answers fast, and generic scripts simply didn't scale.
Challenge: 6 CS agents facing tens of thousands of technical messages a month
As order volume grew, Sidotech's 6-person customer support team couldn't keep up with the AfterSales load arriving simultaneously across Zalo, Facebook Messenger, and Instagram DM. The majority of incoming messages were technically repetitive: how to install drivers, where to activate warranty, what to do when a device isn't recognized.
The problem wasn't just volume. A 100+ SKU catalog made rigid chatbot flows impractical: each keyboard or mouse model had its own firmware version, OS compatibility matrix, and troubleshooting steps. A scripted bot would quickly hit edge cases and either respond incorrectly or fail to respond at all.
Nights and weekends were the hardest gap: customers receiving hardware at 9 PM, hitting a connectivity error, and waiting until the next morning for a reply. Average response time exceeded 4 minutes during peak hours, with zero coverage after 10 PM.
Solution: Knowledge Base + omnichannel AI Agent
A full-catalog Knowledge Base — so AI retrieves the right technical answer instantly
Rather than building rigid conversation trees, Sidotech and Filum built a Knowledge Base covering the entire product catalog: installation guides, warranty activation flows, driver documentation, and per-model troubleshooting tables. Filum's AI Agent was trained to retrieve the right content based on the customer's described symptom and model — no need for the customer to know the exact product name.
AI Agent deployed across all three channels, 24/7
The AI Agent was deployed simultaneously on Zalo OA, Facebook Page, and Instagram — unified through a single Filum Inbox. CS agents no longer had to monitor three separate interfaces, and customers receive a response within seconds regardless of the channel. For the ~90% of conversations the AI handles end-to-end, no human intervention is needed — including at 2 AM.
More complex cases — special warranty claims, out-of-policy returns, escalated complaints — are routed to human agents with full conversation context attached, so agents can respond without asking customers to repeat themselves.
Data-driven continuous improvement
Every month, Filum analyzes the 12 most common request types from live conversation data, helping Sidotech update the Knowledge Base and refine AI behavior based on actual usage patterns. The AI resolution rate grew from over 70% in the first month to ~90% after several improvement cycles.
Results
- ~90% of AfterSales conversations resolved end-to-end by AI — up from 70%+ in month one, driven by iterative Knowledge Base and behavior refinements
- Response time from 4+ minutes to a 10-second target — most visible during evenings and weekends when there was previously zero coverage
- CS team focused on high-value cases — special warranty, upgrade advisory, complex complaints; no longer buried in repetitive technical queries
- >87% task completion rate on order guides and device connectivity troubleshooting flows
"What keeps us with Filum isn't just AI quality — it's how Filum proactively partners with us to improve at every stage. That kind of partnership is rare."
— Truong Nguyen, CEO, Sidotech
Want to achieve similar results for your business?
Explore more success stories at filum.ai/customers.
