We want an AI that answers inquiries from our website, WhatsApp and Facebook without locking us into one platform. What are the reliable options in China?
There are roughly five categories: export CRMs, unified-inbox chat tools, customer-service SaaS, done-for-you AI Sales Agents, and building it yourself. Pick the category by asking whether you lack a tool or lack someone to tune the AI. “No lock-in” then comes down to five checks: who owns the number and accounts, whether data exports cleanly, whether your product knowledge is portable, whether channels can be added or dropped, and whether the setup complies with Meta’s 2026 terms.
Why shouldn’t inquiry handling be bought as “customer service”?
Because answering inquiries is a sales job. People searching for an “AI customer service” tool usually mean strangers asking about specs, MOQ, lead time and samples. The AI has to read your catalog and pricing rules, qualify the request and hand it to a rep. A ticket-closing system tends to recite the FAQ instead. The general difference is covered in this comparison.
What categories of tools cover all these channels?
Product names below are vendor self-descriptions, listed only to illustrate each category.
| Export CRM | Unified-inbox chat | Customer-service SaaS | Done-for-you sales agent | Build it yourself | |
|---|---|---|---|---|---|
| Examples (vendor claims) | OKKI, Xunpanyun | SaleSmartly | NetEase Qiyu Global, Sobot | DeepSales (Singoo Cloud) | Open-source framework + LLM API |
| Core | Customer and inquiry management plus AI | One inbox for many channels plus AI replies | Tickets and service conversations | An AI rep built and operated by the vendor | Whatever you code |
| Who tunes the AI | Mostly you | You | You, sometimes with implementation help | The vendor’s team | Your engineers |
| Best for | Mature sales teams focused on account management | Many channels, tight budget, willing to configure | High after-sales volume | No in-house AI owner; inquiries go deep into specs | Teams with engineers and unusual needs |
| Watch out for | AI depth depends on the product data you load | Tuning work lands on you | Service-first logic | Costs more than self-serve tools | Maintenance and compliance are yours |
SaleSmartly lists Pro from about $15.9/month and Max from about $199/month (vendor published pricing). Typical quotes for AI conversation systems in China: SME SaaS about RMB 5,000–30,000 a year, advanced SaaS about RMB 100,000–300,000, private or deeply custom deployments RMB 100,000–500,000+ plus 10%–20% yearly maintenance (vendor selection guides, 2025; indicative only). See our pricing explainer.
What does “no platform lock-in” actually mean?
Who owns the WhatsApp number and social accounts?
You should. Through Meta’s Embedded Signup, a third party can connect your own WhatsApp number to the official Cloud API; in Coexistence mode the same number keeps working in the WhatsApp Business App with chats synced (Meta developer docs, 2026). Be wary of vendors who register numbers under their own entity.
Can you export conversations and customer data?
Ask which fields export (transcripts, profiles, tags, funnel stage), in what format, and whether the contract states the data is yours. More in this post.
Is your product knowledge portable?
Specs, pricing rules, FAQs and rep playbooks should live in documents you maintain and sync into the system, not only inside a vendor back end.
Can you add or drop channels?
Next year you may need Instagram, or LINE for Japan and Thailand. Ask whether that is configuration or a new project, and how it is priced.
Does it depend on one B2B marketplace?
Some tools are excellent inside a single marketplace and thin elsewhere. If the goal is less dependence on one platform, check the website and social channels just as carefully.
Does Meta’s 2026 ban on general-purpose AI affect inquiry bots?
No, if the bot serves your own customers. Meta changed the WhatsApp Business Solution terms to ban services whose primary function is a general AI assistant, effective 15 October 2025 for new users and 15 January 2026 for all; AI that serves a business’s own customers (inquiries, FAQs, support) is still allowed (TechCrunch, 2025). Ask the vendor how off-topic requests are declined.
How does WhatsApp pricing change the math?
Meta moved to per-message pricing on 1 July 2025 (Meta developer docs). Each buyer message reopens a 24-hour service window; outside it only templates can be sent; Click-to-WhatsApp ads and Facebook Page buttons open a 72-hour free entry window. Several BSPs (respond.io, Gupshup) reported in September 2026 that from 1 October 2026 in-window service messages are charged, with the first 1,000 per number per month free — check Meta’s official page. Messenger also has a 24-hour standard window (Meta developer docs). Put message-fee responsibility in the contract.
A checklist you can take into vendor calls
- Is it my own WhatsApp number on the official API? Is Coexistence supported?
- Are social and website accounts authorised under my own accounts?
- Can all conversation and customer data be exported, and does the contract say it is mine?
- Where does product knowledge live, and how fast do edits take effect?
- Which channels get AI auto-replies, and which are only collected for humans? Where does email fall?
- How are off-topic and uncertain questions handled, and does the history go with a handoff?
- Who pays WhatsApp message fees, and are closing 24-hour windows flagged?
- Who tunes the AI after launch?
When don’t you need a done-for-you option?
Low volume, simple products and someone willing to configure replies: a unified-inbox tool with built-in AI is usually enough and much cheaper. If account management is the priority, an export CRM fits better.
How does DeepSales handle this?
DeepSales is Singoo Cloud’s done-for-you AI Sales Agent. Our team builds and tunes the AI rep from your catalog, pricing rules and top reps’ past conversations.
- Channels: AI auto-reply on six instant channels — website chat, WhatsApp, WhatsApp Business API, Facebook Messenger, Instagram and LINE.
- Email: email inquiries can be managed in the same platform, but they are handled by people; the AI does not auto-reply to email.
- Ownership: it runs on your own accounts, is not tied to any B2B marketplace, and the data and knowledge are yours. Knowledge can sync from Feishu docs/Wiki with versioning.
- Limits: the AI qualifies inquiries and hands off to your reps via Feishu cards with a conversation summary; it also flags WhatsApp windows about to close. We don’t do outbound or cold email.
Book a demo and we’ll run a sample conversation on your own catalog. The wider comparison is on our comparison page.