AI transparency

In short

Autoflowly builds and runs business apps with AI. Under the EU AI Act (Regulation (EU) 2024/1689, Article 50), a person talking to an AI system must be told so. We do this by default, wherever our AI talks to people, in their language.

• Every AI that talks to people is labelled as an AI and is instructed never to claim to be a person.
• When someone asks for a person, the AI offers to hand over to the business or to the Autoflowly team.
• What reaches a business's customers or moves money waits for the owner's approval — except automatic replies and the phone assistant, which the owner switches on and which say they are an AI.

Where AI is used, and how people are told

• Flo, Autoflowly's support assistant (in the app): labelled “AI assistant”. What it cannot settle — refunds, account security, “I want a person” — goes to the Autoflowly team as a ticket.
• Flo on autoflowly.com (for visitors): labelled as an AI assistant. You can delete your chat history at any time with the bin icon in the chat. Chats are deleted after 6 months, or after 2 years if you left your email to be contacted.
• The app builder: AI writes the app from your description. You review what is built; the app and its content are yours.
• The AI receptionist in apps built on Autoflowly (phone, SMS, WhatsApp, website messages): calls open with “I'm …, the AI assistant”, and the first automatic message of a conversation says it is an automatic reply from the business's AI assistant and that the team can take over.
• AI agents published on a business's website: the chat window is labelled “AI assistant” in the visitor's language.
• The AI team working for a business owner: agents prepare work — replies, reminders, follow-ups. Anything that reaches a customer or touches money waits for the owner's approval in the Inbox, unless the owner has switched on automatic replies (labelled, as above).

Human oversight

The business owner decides what the AI may do on its own: automatic replies and the phone assistant are off until the owner switches them on, and other customer-facing actions are approved first. Every conversation stays in the owner's Inbox, where a person can take over at any time.

AI models and providers

Autoflowly uses general-purpose AI models from OpenAI, Anthropic and Google (Gemini), chosen per task. Voice commands are transcribed with OpenAI Whisper; phone calls run through Twilio, which recognises speech and speaks with Google voices. A business can connect its own OpenAI or Anthropic key. The obligations for the general-purpose models themselves lie with their providers.

What we do not do

• No emotion recognition, no biometric identification or categorisation, no social scoring, no manipulative or deceptive techniques.
• Our ready-made apps do not let AI decide on hiring, credit, insurance or students' grades. Those decisions stay with people.

If your business uses Autoflowly

Under the AI Act, your business is the deployer of the AI in your app. Autoflowly labels it for you. Please:

• Do not remove or hide the AI labels and notices, and keep a person reachable for customers who ask for one.
• Do not instruct the AI to pretend to be a person — it is told to refuse.
• If you plan to use AI for a high-risk purpose — screening job applicants, judging creditworthiness, evaluating students, or deciding access to essential services — contact us first: those uses carry extra obligations (Articles 26 and 27 of the AI Act) that our ready-made apps do not cover.
• If you publish AI-generated images, audio or video that look real, say that they are AI-generated.
• Make sure the people who use the AI in your business understand what it does and where its limits are (Article 4). Flo can explain any part of Autoflowly's AI.

Limits

AI can be wrong. Autoflowly's assistants answer from the business's own data and FAQ where they can, and are told not to invent prices, bookings or policies — still, check anything important before relying on it.