AI follow-up on the phone, with a human at the controls
Social Agent App puts the AI follow-up of a social selling platform on the phone. Every draft the AI is not allowed to handle on its own waits as a card until someone reads it and swipes. In the App Store, in the brand of the agency that signs in.
- Art. 50
- AI disclosure, enforced by the platform
- 1
- swipe to send or archive an AI draft once it has been read
- iOS
- in the App Store, in the brand of the agency that signs in
The challenge
An agency doing social selling for its clients lets the AI do most of the work: finding leads on LinkedIn, inviting, following up. But whoever wants to keep an eye on that from the car or between two meetings does not open a portal. And yet every reply that goes out in a client's name has to pass someone who signs off on it.
Autonomous where it can be, a human where it must be. At the other end of every message is a real person who has the right to know they are dealing with AI. Since August 2026 that is no longer etiquette but an enforceable obligation under Article 50 of the EU AI Act.
The solution
Social Agent App is the mobile front end of the platform the campaigns run on. The app runs no campaigns of its own; it puts the approval on the phone.
Each card shows what the lead wrote and what the AI wants to say back. Swipe right to send, swipe left to archive, tap to open the draft and rewrite it first. Phone numbers, addresses and links in the lead's message are tappable. Whoever thinks in a different language than the lead writes in their own; the app sends in the lead's. Next to the draft sits the context: who it is, where they work and what has already been said. A booking link goes in with one tap.
Replies the AI cannot place are not guessed at: they land with a human as an empty card.
The line between AI and human is a setting, not a promise
Each campaign has one slider. Left of the knob the AI handles things itself, everything to the right comes in as a card for approval. All the way left the AI does nothing on its own, all the way right everything. The recommended position lets the AI handle only the rejections, and a meeting can only be booked automatically once the calendar is connected.
That makes human oversight not a policy rule but a property of the product. That is exactly what we mean by compliance as code. The same goes for the transparency obligation: the Article 50 AI disclosure sits in the platform as a mandatory, non-bypassable field. No setting you can switch off, no checkbox that happens to be ticked by default: a message without the disclosure does not get through.
In the App Store
Social Agent App has been in the App Store since September 2026, published by Second Company. It is a white-label app: once signed in, it takes on the agency's colours and logo, so a team sees its own brand and not ours. Bug fixes and small improvements reach the phone as over-the-air updates, without a new review; only a change to the native layer needs a new build.
Results
- Human oversight as a property of the product: the AI writes, a human decides with one swipe, and per campaign it is fixed where that line lies
- Demonstrably compliant with the EU AI Act transparency obligation: the AI disclosure lives in the platform, not in a manual
- In the App Store since September 2026, white-label: once signed in, the app wears the agency's colours and logo
Related service
EU AI Act & compliance
Lawyers tell you what has to happen. We make sure it actually does: in the code, in the pipeline, in production. Transparency obligations, human oversight, logging and documentation: we build it in from day one, so compliance is not a brake but a given.
Request an AI Act quick scanNext case
An AI agent that rewrites job ads and publishes them across 600 platforms
Brockmeyer · Recruitment & job marketing
Read the case