AI Transparency

Scannect currently uses AI in three places:

• Business-card extraction. When you scan a business card, an AI vision model reads the image and returns structured fields (name, title, company, phone, email, and similar) that you can edit before saving.
• AI email drafts. When you request a draft, an AI model returns a plaintext email tailored to the selected contact and template.
• Voice-note transcription. When you record a voice note on a contact, an AI speech-to-text model returns the transcript, which you can edit before saving as a note. Recordings are capped at 120 seconds and are not retained after transcription completes.

All AI requests — including chat/text and voice-note transcription — are made from Scannect's server to the Lovable AI Gateway. Your browser does not call the models or the underlying providers directly. As of the review date on this page (July 16, 2026), the chat/text model used for both extraction and drafting is a Google Gemini preview model routed through the gateway, and voice-note transcription is routed to the gateway endpoint /v1/audio/transcriptions using model openai/gpt-4o-mini-transcribe. The specific models, providers, and gateway configuration may change over time; this page is updated when they do.

Depending on which feature you use, an AI request may include:

• the business-card image you uploaded and prompts describing the extraction task;
• for email drafts, the selected contact fields, the selected template, your saved sender identity, business profile, and signature from Settings, and the drafting instructions you provide;
• for voice notes, the raw WAV audio you recorded (up to 120 seconds) sent through the Lovable AI Gateway to the transcription model. The audio is used only to produce the transcript and is not written to Scannect storage; the transcript itself is saved only when you explicitly save the note on a contact;
• a system prompt that describes the task and requested output format.

Do not include any information in your business profile, settings, notes, drafting instructions, or spoken voice notes that you would not want processed by an AI provider.

PHI / patient-identifying information. Scannect is a sales CRM, not a HIPAA-covered system. Do not speak, type, or upload patient names, dates of birth, medical record numbers, diagnoses, treatments, images, or any other patient-identifying information into card scans, notes, drafts, or voice recordings. Content you submit is transmitted to third-party AI providers via the Lovable AI Gateway for processing.

The API credentials used to call the AI gateway are held only on the server. They are not shipped to the browser and are not exposed to Customer Data or to connected accounts.

AI output can be inaccurate, incomplete, biased, or otherwise unsuitable. You are responsible for reviewing, editing, and approving extracted fields and email drafts before saving or sending. Scannect does not automatically send AI-generated emails on your behalf without your action.

AI output from Scannect is not medical, legal, regulatory, financial, tax, or compliance advice, and must not be used for clinical decision-making. Consult qualified professionals for those matters.

Do not submit patient records, protected health information (PHI), diagnoses, treatment or prescription data, laboratory results, medical images, or insurance/claims data to any Scannect feature, including AI features. See our Terms.

Scannect does not intentionally use Customer Data to train a generally available model. Third-party processing of AI inputs and outputs by the gateway and the underlying model provider is governed by those providers' agreements and account settings. We do not claim provider zero-retention unless a specific provider contract or setting we have verified says so.

Card scans, AI drafts, and voice notes are metered per workspace and per plan. Server-side checks enforce those limits before any request is sent to an AI provider, and voice recordings are additionally capped at 120 seconds per note. See Pricing for current limits.

We may change AI providers, models, prompts, or safety configurations over time. When we do, this page will be updated. Questions? Contact support@myscannect.com.