Social Media Agents — cover
Back to Projects

Social Media Agents

A Social Media Post Agent that takes Atolls from a Slack message to a published LinkedIn post, with a human approving everything that goes live.

Role
Agents · N8N
Duration
1 semester
Semester
WS 2025/26
Year
2025/26
01

Context

Social Media Post Agent is a team project built for Atolls, formerly known as Global Savings Group. Atolls is a privately held international commerce content company working in online savings, cashback and affiliate marketing, combining multiple savings portals across more than 20 countries. It builds and runs digital shopping destinations that help millions of consumers make smart spending decisions. A company at that scale posts constantly on LinkedIn, so the goal of this project was to take content creation all the way from a request in Slack to a published LinkedIn post automatically, while keeping a human in control of what actually goes live.

02

Problem

Creating content is easy. Scaling it is not. Atolls posts many different kinds of content, from recruitment to product showcases to employee engagement, and each one costs time, has to stay consistent with the brand voice, and needs sign off before it goes out. Done by hand every week the workflow is slow and repetitive, and one rushed post can easily end up off brand.

03

Solution

A content automation system that runs from Slack to LinkedIn. A team member triggers a request in Slack, an intelligent router sends it to the right specialised agent (recruitment, product showcase or employee engagement), and the agent drafts a post in a structured, predictable format. The draft comes back into Slack for human review, where it can be approved, edited, regenerated or rejected, and only approved posts get published. The system is built on n8n with Airtable as a single source of truth, and it separates content generation from the approval process so it stays maintainable and reliable. The result: repetition is automated while strategy stays human, with faster production, consistent branding, and a setup that scales across multiple post types in parallel.

Take a look at the result

01

Modular Workflows

Each post type is isolated, so the system stays easy to maintain and extend.

02

Structured Outputs

Posts are generated in a JSON format that is predictable, parseable and reliable.

03

Human in the Loop

Approval is not optional, it is built into the architecture.

04

Intelligent Routing

Social Media Agents — Intelligent Routing
Routing

An AI router reads each request and sends it to the right specialised agent: recruitment, product showcase or employee engagement. Each agent keeps its own memory, so it holds the tone and context of its content type.

05

Two-Workflow Separation

Architecture

Content generation and the approval process run as two separate workflows. Keeping them apart avoids fragile, spaghetti style automation and makes the whole system easier to debug and maintain.

Social Media Agents — Two-Workflow Separation
06

Airtable as a Single Source

Social Media Agents — Airtable as a Single Source
Data

Airtable holds every post and its status in one place, from pending text to published. It acts as the single source of truth that both workflows read from and write to.

07

Repetition Automated, Strategy Stays Human

One flow replaces multiple handovers for faster production. A templated structure with human approval keeps the brand consistent. The setup scales across post types in parallel, and the AI generates while people decide.

08

Delivered & Handed Over

The system was finished and handed over to Atolls, where it now runs in production. After handover the agents keep learning from new content, and the next unlock is extending the system to more channels and post types.

Tools & Technologies
n8nn8n
+
SlackSlack
+
AirtableAirtable
+
LinkedInLinkedIn