Virtual AI customer service employee for ecommerce brands
Hire an AI support employee that answers every customer email
AI Auto Support takes over your brand inbox, learns from historical emails, website FAQ, and Shopify policies, then drafts replies in the customer language. You approve and instruct — swipe on your phone, tap a button in team chat, and delegated low-risk replies send themselves.
Approve anywhere
Phone card swipes, Lark/DingTalk/WeCom cards, desktop pet
90-day mailbox learning
Frequent issues and reply boundaries
Policy trust audit
Return, shipping, warranty facts need review
Open MCP ecosystem
15 standard tools for your own AI agents
A day in your AI employee's life
It works like an employee, not software waiting for clicks
Morning · Inbox takeover
Syncs the mailbox and moves support emails into one AI Support folder; everything else stays untouched.
Daytime · Drafting
Detects the issue, looks up the Shopify order, cites audited knowledge, and drafts in the customer language.
Anytime · Your review
Cases that need you go to your phone and team chat; swipe right to approve or reply in Chinese to redirect.
Delegated · Auto-send
Low-risk replies within your delegation send themselves, with a 15-second undo window.
Evening · Daily report
Reports how many emails were handled, auto-sent, and intercepted today.
Core capabilities
Email takeover, not a draft widget
AI imports support emails, marks them read, and moves them to one AI Support folder while leaving other mail untouched.
Knowledge bootstrap + policy trust audit
90 days of email history, website FAQ, product and policy pages become knowledge candidates; return/shipping/warranty facts are audited before use.
Supervise in Chinese, reply in the customer language
Chinese summaries and review drafts for every case; type or hold-to-talk in Chinese and AI regenerates the final customer-language reply.
Delegated, safe auto-reply
Delegate automation per email type with card budgets and quiet hours; refunds, replacements, complaints, and chargebacks always need a human.
Approvals where you already are
Swipe cards on your phone, tap buttons in Lark/DingTalk/WeCom, or let the desktop pet raise a sign; 15-second undo after approval.
Open agent ecosystem
Agent Gateway REST API plus 15 MCP tools connect approvals and instructions to your own Claude / ChatGPT / self-hosted agents.
Chinese-first workspace
Customers write in any language — you review in Chinese
Every inbound email gets a Chinese summary, and every AI draft ships with a Chinese review copy plus its knowledge sources. Approve, correct, or delegate from one queue while the AI replies in the customer's language. Below is the real review workspace.
Approval surfaces
Approve without sitting at a desk
Traditional helpdesks keep you in a web dashboard. AI Auto Support pushes decisions to the places you already are.
Card swipes on your phone
A card deck: swipe right to approve, left to reject, up to snooze, and hold to talk to instruct in Chinese.
Lark / DingTalk / WeCom
Approval cards land in your team chat; tap a button to approve, and any reply becomes an instruction to the AI.
Desktop pet
A little robot works quietly in the corner of your desktop and raises a sign only when it needs you.
15-second undo window
Every approval has a 15-second undo countdown, and high-risk actions require double confirmation.
How it differs from helpdesk tools
Helpdesk + AI add-on
AI Auto Support
Positioning
Helpdesk + AI add-on ·You run the ticketing software; AI only suggests
AI Auto Support ·An AI employee runs the full loop: triage, order lookup, drafting, sending, knowledge
Management language
Helpdesk + AI add-on ·English dashboard, editing English emails line by line
AI Auto Support ·Chinese summaries and instructions; AI writes the customer-language reply
Approvals
Helpdesk + AI add-on ·Log into a web dashboard
AI Auto Support ·Phone card swipes, Lark/DingTalk/WeCom cards, desktop pet
Knowledge & safety
Helpdesk + AI add-on ·Answers from whatever was crawled
AI Auto Support ·Policies pass a trust audit; low-confidence knowledge never auto-replies; high risk needs a human
Pricing
Helpdesk + AI add-on ·Per-seat fees plus per-AI-resolution fees
AI Auto Support ·No seats: priced by brand AI employees, email volume, and AI usage; credits never expire
Ecosystem
Helpdesk + AI add-on ·Closed black box
AI Auto Support ·Agent Gateway + MCP open protocol for your own AI agents
Why brands must stay isolated
One brand needs one isolated support brain
If multiple brands share one AI, knowledge, policies, tone, and automation limits become unsafe. One account can manage multiple brands, but every brand needs its own workspace, knowledge base, inbox takeover, and automation policy.
- Same-brand website, FAQ, Shopify, Amazon, and team chats can be knowledge sources for one AI.
- Different brands must become separate AI employees and cannot share reply policy.
- Packaging charges by brand AI employees, email volume, AI usage, and advanced integrations — never by seats.
Launch workflow
Connect inbox
Connect IMAP/SMTP and create or reuse one AI Support folder.
Learn context
Backfill mailbox history and crawl website, FAQ, and policy pages.
Confirm boundaries
Approve policies, tone, risk rules, and delegation scope in Chinese.
Start working
AI handles email while you approve from your phone and team chat; delegated replies auto-send, with a daily report.
Testing-period packages
Starter
¥99/mo + AI usage
For new Shopify DTC sellers — auto-reply is included from the first tier.
- AI email takeover
- Supervise the AI in Chinese
- Auto-reply
- Basic mailbox history learning
Professional
¥199/mo + AI usage
For stores with steady support volume — multiple inboxes, rules, and knowledge conflict checks.
- Higher email quota
- Knowledge conflict detection
- Daily AI support report
- More team seats
Peace of Mind
¥499/mo + AI usage
For high-volume single brands that need stronger automation and auditing.
- Dedicated brand knowledge base
- Advanced auto-reply policies
- Auto-reply audit trail
- Team permission management
Custom
Contact sales
For high email volume, multi-brand groups, strict privacy, or dedicated deployment — each brand still runs its own AI agent.
- Isolated database
- Dedicated deployment
- Custom ERP/WMS/logistics integrations
- Hands-on onboarding service
FAQ
Make the core loop safe first, then delegate
The launch priority is inbox takeover, knowledge bootstrap, policy audit, and delegated auto-reply — automation scope is approved type by type, never switched on all at once.
Why not one AI for multiple brands?
The agent learns tone, policies, historical handling, and automation boundaries. Mixing brands creates a direct wrong-reply risk, so every brand gets its own AI employee.
Is auto-reply safe?
Automation is delegated per email type; refunds, replacements, complaints, and chargebacks always need a human. Card budgets, quiet hours, and a 15-second undo window add extra guardrails.
My English isn't great — can I still run overseas support?
Yes. Every case has a Chinese summary and review draft. You instruct the AI in Chinese by typing or holding to talk, and it writes the final reply in the customer language.
How is this different from Zendesk or Gorgias?
Those are ticketing systems with an AI copilot, billed per seat plus per AI resolution, and you still work an English dashboard. This is a full AI employee: it handles, you approve — with no per-seat fees.
Can I plug in my own AI tools?
Yes. Agent Gateway exposes a REST API and 15 MCP tools, so Claude Code, ChatGPT, or self-hosted agents can approve and instruct.
Put your AI employee to work today
Connect your inbox and it starts learning your brand immediately — shadow-run first, then delegate auto-reply type by type.