Drowning in "Where's My Order?"
A direct-to-consumer skincare brand in Brighton was growing fast — which meant their two-person customer support team was drowning. Order status queries, returns requests, product questions, shipping updates. The same types of questions, hundreds of times a week, across email, WhatsApp, and their website chat widget.
"We were spending so much time on repetitive queries that we couldn't give proper attention to the customers who actually needed help," the operations manager said. "The complex issues — allergies, damaged products, subscription changes — those are the ones that need a human. But we were too buried in 'where's my order?' to get to them."
The Multi-Channel Agent
We deployed OpenClaw on a VPS and connected it to three channels simultaneously:
- WhatsApp Business — customers can message the brand directly and get instant responses about orders, ingredients, and shipping.
- Email — the agent monitors the support inbox, categorises incoming messages, and handles straightforward queries with personalised responses.
- Webchat — embedded on the brand's website, offering immediate assistance during the browsing and checkout process.
The agent connects to Shopify for real-time order data, the shipping carrier's API for tracking, and the brand's product knowledge base for ingredient and usage information.
The Memory Advantage
The feature that surprised the team most was OpenClaw's persistent memory. When a returning customer messages, the agent remembers their previous interactions, order history, and any past issues.
"A customer messaged about a rash from a new product. The agent already knew she'd had a sensitivity issue six months ago with a different product, pulled up both orders, and flagged it for human review with full context. Our support team would have had to dig through three different systems to piece that together."
The Numbers
After the first month, 65% of incoming support queries were being resolved entirely by the agent — no human intervention needed. The remaining 35% were escalated to the human team with full context already attached: customer history, order details, and a suggested resolution.
Response times dropped from an average of 4 hours to under 2 minutes for agent-handled queries. Customer satisfaction scores actually went up — customers don't care whether a human or an AI answers their question, as long as the answer is fast and correct.
The two-person support team? They're still two people. But now they're handling the work that actually requires empathy, judgement, and problem-solving — rather than copy-pasting tracking numbers.