Skip to content
ZhaiBro

Trust Center

ZhaiBro Trust Center.

ZhaiBro is the B2B SaaS support operations product offered at zhaibro.com. This page explains its AI workflow, data boundaries, security controls, and human-review requirements.

WorkspaceApproved knowledgeAWS BedrockHuman review
MODEL USE

AWS Bedrock use

ZhaiBro uses Anthropic Claude through Amazon Bedrock to classify SaaS support requests and generate cited support drafts for human review.

  1. 01Browser clients never receive Bedrock credentials or model provider keys.
  2. 02Classification requests include the support subject and customer message. Draft requests include the customer message and relevant excerpts retrieved from approved workspace knowledge.
  3. 03Draft instructions restrict the model to provided company knowledge and prohibit unsupported refund or account-action promises.
  4. 04Invalid model responses fall back to deterministic local processing; generated replies remain editable drafts and are not sent automatically.
  5. 05Usage events record provider, model, token estimate, workspace, and audit context for reviewability.

Live reviewer evidence

Current operational review state.

Status values come from the ZhaiBro backend. Unavailable evidence is never presented as ready.

business_purpose

Business purpose

ZhaiBro is a B2B SaaS support operations workspace for small SaaS teams, focused on approved knowledge, support triage, cited AI drafts, and human review.

Ready
aws_bedrock

AWS Bedrock use

AWS Bedrock Claude is used server-side for support ticket classification and draft generation when production credentials are configured; browser clients never receive model credentials.

Ready
human_review

Human review boundary

AI drafts stay inside the workspace until a human reviewer edits, approves, resolves, or copies the response. The MVP does not automatically send customer replies.

Ready
tenant_isolation

Tenant isolation

Tenant-owned data is scoped by workspace_id in Postgres, and server-side access derives from a revocable session membership check or a workspace-bound API key hash.

Ready
data_deletion

Data deletion controls

Workspace owners can delete imported knowledge sources, including stored text, chunks, and embeddings tied to that source.

Ready
non_financial_scope

Non-financial scope

ZhaiBro sells subscription access to B2B SaaS software. It is not a bank, card issuer, money transmitter, lender, stored-value product, payment facilitator, or virtual card product.

Ready
01

Business purpose

ZhaiBro helps small SaaS teams turn approved internal documentation and customer support questions into human-reviewed AI support drafts.

  • The product imports selected Notion or Google Docs knowledge sources into a workspace knowledge base.
  • Support questions are classified, matched to same-tenant knowledge, and turned into cited draft replies.
  • ZhaiBro does not automatically send customer replies; users review and resolve drafts inside the workspace.
02

Commercial boundary

ZhaiBro is B2B SaaS software. Self-service payment collection is not currently enabled on the website.

  • ZhaiBro is not a bank, payment processor, card issuer, money transmitter, lender, stored-value provider, or financial product.
  • ZhaiBro does not issue virtual cards, hold customer funds, facilitate payments between third parties, or provide money movement services.
  • Any future billing workflow will relate only to access to the ZhaiBro software product.
03

Tenant and data controls

Tenant data is scoped by workspace_id in Postgres, and server-side authorization derives access from session-backed workspace membership or workspace-bound API credentials.

  • Imported document text, chunks, embeddings, tickets, AI drafts, usage events, and audit logs are workspace-scoped.
  • OAuth tokens are encrypted before storage, and API keys are stored only as hashes.
  • Workspace owners can delete imported sources, including text, chunks, and embeddings tied to that workspace source.
04

Responsible use boundary

ZhaiBro is designed for internal support drafting, not autonomous customer communication or high-impact decision making.

  • The service may not be used for medical, legal, financial, employment, housing, insurance, or other regulated decisions.
  • Workspace reviewers remain responsible for checking facts, citations, tone, and customer-specific commitments before using a draft.
  • Suspected abuse or security issues can be reported to [email protected] or [email protected].