AI Disclosure
Effective 2 June 2026 · Updated 8 June 2026
HAMANI CRM ships with HAMANI CONCIERGE — an agentic AI assistant embedded across the product. This page tells you exactly how it works, what we share with our AI vendor, and how to limit it.
1. AI models and vendors
HAMANI CONCIERGE is a connected team of specialist agents. Most are powered by Anthropic Claude (anthropic.com). Some analytical agents (e.g. Insights, Reporter, Reviewer) run on Google Vertex AI (Gemini), and voice features use Google Cloud Speech— both in the Sydney (australia-southeast1) region. HAMANI is not an AI lab; we orchestrate these vendors’ models behind our own product surface, tools, and safety policies. We make no “proprietary HAMANI AI” claims — the providers are named honestly.
2. What we send to the model
When you invoke HAMANI CONCIERGE, the following may be sent to Anthropic:
- The chat messages you type
- Relevant CRM records explicitly retrieved to answer your question (e.g. a deal, a contact, a thread of emails) — scoped to your tenant
- Tool definitions describing what CONCIERGE may do
We do notsend: your entire tenant database, other customers’ data, credentials, payment instruments, or content you have not invoked CONCIERGE on.
3. What Anthropic does with it
Anthropic processes inputs to generate model responses. Per Anthropic’s commercial terms, your inputs and outputs are not used to train Anthropic’s models. HAMANI may retain logs for debugging, abuse prevention, and product improvement, scoped to your tenant.
4. Inaccuracy and limits
Large language models can be wrong, including confidently. Treat CONCIERGE outputs as a draft, not a source of truth. Do not rely on CONCIERGE for legal, medical, financial, or regulated decisions.
5. Tools and actions
HAMANI CONCIERGE may take actions in your workspace via tool calls (e.g. drafting an email, creating a task). Sensitive or destructive actions require your explicit confirmation. Audit logs are visible to workspace admins.
6. Opt-out and configuration
Workspace admins can disable HAMANI CONCIERGE at the tenant level. We will not invoke the model on your tenant data when CONCIERGE is disabled.
7. Model Context Protocol
HAMANI CONCIERGE uses the open Model Context Protocol (MCP) to connect to data sources and tools. This makes auditing and extension safer than ad-hoc plugin architectures.
8. Voice transcription (field staff)
Where enabled, the field-staff app can transcribe a spoken note into a draft progress note using HAMANI CONCIERGE. This feature is opt-in and off by default; audio is processed to produce the text draft, which the worker reviews and edits before it is saved. Drafts are stored as part of the participant’s records under the same privacy and retention terms as typed notes.
9. AI agents and what “ready now” means
HAMANI CONCIERGE is a connected team of specialist agents — including Concierge, Invoicer, Outreach, Pipeline, Goodwill, Privacy Guardian, Contact, Booking, Reviewer, and Coach, plus analytics agents (Insights, Reporter). The live count and names shown on our website are read directly from our agent registry, so they are always accurate; we never describe a roadmap or “building” agent as shipped, and agents marked “building” (e.g. Ambient voice, Forms Workflow) are honestly flagged as awaiting a dependency.
Their guardrails are deliberate and enforced in code:
- Agents never auto-send communications — drafts are always reviewed and sent by you.
- Agents never invent client data; when unsure they say so and show low confidence rather than guess.
- The Goodwill discount is additive only — kindness can earn a small thank-you, but rudeness is never penalised.
- Answers carry a confidence indicator so you can see how sure the agent is.
Our NDIS reform features are described as “ready now” because they are built and live to the published government position. This does notmean the software autonomously rewrites itself. Reform updates are reviewed and shipped by HAMANI’s team as official detail is finalised; our AI agents assist users with their work but do not silently modify the product, your configuration, or your compliance posture without your action.
10. Questions
Email enquiries@hamanicrm.com.au for AI governance, model-card, or due-diligence questions.