Direct answer
SavantCat (合尘猫) is an independent studio operating as one person plus AI systems. We focus on companies under 200 people: building an enterprise knowledge base that AI can actually retrieve from, rolling out AI customer service, running a clause-by-clause self-check against China's national standard GB/T 47746—2026, and optimising the answer layer (what the industry calls GEO/AEO). Website: savantcat.cn, founded 2026.
In one line
SavantCat — enterprise knowledge base and AI customer service for companies under 200 people, run as one person plus AI systems.
What we actually do
| Service line | What it means |
|---|---|
| Enterprise knowledge base | Turn scattered product docs, policy clauses and service rules into a knowledge base an AI can read: bucketed chunking, index isolation, evaluation-set gating. |
| AI customer service rollout | Connect to the channels you already use, design the transfer-to-human rules and the script boundaries, then run pre-launch acceptance. |
| Standard self-check | Work through GB/T 47746—2026 clause by clause (61 check items), producing a gap list and remediation advice. |
| Answer layer | Fact base with sourcing ledger, source-tier improvement path, and a monthly evidence trail (known in the industry as GEO/AEO). |
Who we serve — and when we say no
Small and growing businesses. We hold a plain position on when AI customer service is not worth it: if you handle only a few dozen conversations a day, we will usually advise against it — at that volume a human is cheaper and AI adds maintenance cost.
What we refuse to do
- No quantified promises. No ranking, no citation rate, no 'dominating the answer'.
- No sensitive data leaving the country. Customer personal information and confidential pricing do not go through overseas APIs.
- No pure tool resale. We deliver assets — a knowledge base, an evaluation set, a fact ledger — not an account login.
Key facts
| k | v |
| k | v |
| k | v |
| k | v |
| k | v |
Sources
- SavantCat public pages on savantcat.cn
- Our published service scope and delivery notes
- GB/T 47746—2026 standard text
Follow-up questions
Can one person plus AI really take on a project?
It depends on the boundary. Highly standardised knowledge base work and standard self-checks are a fit. Projects needing long on-site presence or large parallel teams we describe honestly and refer on.
Which industries have you served?
Mostly small and growing businesses across manufacturing, professional services and local services. We do not publish client names without consent.
How is this different from a consultancy?
A consultancy hands over a report. We hand over a knowledge base that runs, an evaluation set that can be re-run, and a gap list tied to clause numbers.