Atmitli Privacy

1. What data we handle

Depending on the surfaces you enable, we process: (a) audio recordings from discovery and department-head interviews; (b) transcripts generated from that audio; (c) business documents and knowledge you upload (procedures, listings, price lists, FAQs); (d) AI-generated drafts and the graded records of those drafts; (e) contact and lead details submitted through this website (name, email, phone, company, message); and (f) account identifiers needed to sign you in.

We do not sell data, and we do not use your business data to train third-party foundation models.

2. Sub-processors (who else touches your data)

Clerk — authentication and account identity.

Deepgram — speech-to-text transcription of interview audio.

Anthropic (Claude) — the AI model that reads your knowledge base and produces drafts.

Neon — managed PostgreSQL database for account and operational records.

Cloudflare — DNS, CDN, and email routing for hello@atmitli.com.

Vercel — hosting for this public website.

Hostinger — hosting for the application server (VPS) that runs your agents.

Each sub-processor receives only the data needed to perform its function. When a new sub-processor is added (for example, a WhatsApp gateway host when messaging goes live), this list is updated before that processing begins.

3. Where your data lives (residency)

The application server that runs your agents and stores your operational data is hosted with Hostinger in a data center in Paris, France (European Union). The managed database (Neon) and website host (Vercel) operate from their own regions.

For regulated organizations that require every byte to remain on their own infrastructure, the Client-Server deployment tier runs the entire intelligence layer on servers you control — see /security.

4. How data flows to AI providers (flow-through)

Interview audio is sent to Deepgram to produce a transcript, then the audio is retained on an encrypted volume and auto-deleted after 90 days.

To produce a draft, the relevant transcript, knowledge-base excerpts, and conversation context are sent to Anthropic (Claude). Anthropic processes this input to generate the draft and, per its API terms, does not use it to train its models. Only the data needed for the specific task leaves your tenant; it is not shared with other clients.

5. Draft storage disclosure

To make quality reviewable and demos honest, Atmitli stores the full text of AI-generated drafts — both the ones approved and the ones the quality harness rejected — in your own client directory on the application server. These draft bodies are not committed to source control and are accessible only to your workspace and to Atmitli operations for support and quality assurance.

6. Retention

Interview audio: auto-deleted after 90 days. Transcripts, knowledge base, and drafts: retained for the life of your account so your agents keep improving, and deleted on written request or account closure. Website lead submissions: retained until the enquiry is resolved and then archived.

7. Your rights and contact

You can request access to, correction of, or deletion of your data at any time. Email hello@atmitli.com and we will action the request and confirm when it is done.