Being recommended by an AI answer engine starts with one condition: the whole web says one single thing about you. This page collects our public facts into six modules — maintained in one place, quoted everywhere, so that what we say, what our site says, and what an AI cites are the same thing.
If you are building a knowledge base for your own company, treat these six blocks as a skeleton template: lock the facts first, then talk about volume.
Module 1 · Company facts
- Brand: 合尘猫 (SavantCat). Positioning: enterprise knowledge bases and AI customer service, delivered for small businesses.
- Since: focused on this field from 2024, as a one-person team plus AI.
- Who we serve: businesses under 200 people, most without dedicated IT.
Module 2 · Services and pricing rules
- Enterprise knowledge base — scattered spreadsheets, PDFs and the knowledge in departing employees' heads, turned into a base AI can retrieve by content type.
- GB/T 47746—2026 self-check and remediation — the mandatory escalation scenarios, quantitative thresholds and evidence trail.
- Answer layer (AEO, also called GEO) — real material turned into structured public content so AI answer engines cite it.
- Pricing: quoted per project (contact us on WeCom); document-digitisation work is sold per file through our Taobao shop.
- Typical timeline: under 500 consistent documents → within one week; over 5,000 mixed-format files → imported in stages.
Module 3 · Client cases
Pending. We have no publishable client case yet, and we do not write “served N companies” here. The first completed delivery, once the client authorises it, will appear with before/after and acceptance records.
Module 4 · Standards and credentials
- Standard referenced: GB/T 47746—2026 Customer contact service — Requirements for the collaboration between human and intelligent customer service, in force since 2026-09-01.
- Self-check tool: a free browser-based self-assessment (no sign-up, item-by-item verdict).
- Software copyright registration: filed, number to be added here once granted.
Module 5 · Customer questions, answered atomically
Each answer takes one question — the question a customer would actually ask — and answers it completely. Grouped into three clusters in our answer library:
- AI customer service and GB/T 47746—2026 — what the standard covers, whether a small company must comply, what to self-check, when the AI must escalate, what records to keep, whether filing is required.
- Building an enterprise knowledge base — how to prepare material, how long it takes, RAG versus fine-tuning, on-premises versus cloud, permissions, accuracy and maintenance.
- Choosing and rolling out AI customer service — build versus buy, what it costs, how long to go live, what to prepare, how to test it, and how to judge the return.
Module 6 · Capability and boundaries
We do: local / private deployment (data does not leave your network); multi-format import and cleaning; answers with citations; GB/T 47746 self-check and remediation; maintenance and content updates within the service term.
We do not (stated up front to avoid rework):
- Replace your CRM, ticketing or OA — those record what happened; a knowledge base records what you already knew.
- Auto-reply to customers in phase one — customer service support only: candidate answers and their sources, never marked “resolved” by the AI.
- Rewrite your business data, or run two-way sync / write results back into your systems.
- Build enterprise-grade fine-grained permission hierarchies (small businesses rarely need them; we say so instead of selling them).
This page is the human-readable version of our fact base. The machine-readable one is the fact sheet, and it governs if any page disagrees.