Home · Services · Use Cases · Industries · Case Studies · Products · About · Book a consultation · Contact
Home / Case studies

Social media automation for agencies across multiple brands

A single system that generates and publishes content for several businesses across three platforms, with optional human approval.

The problem

Creating, scheduling and posting content for several businesses across three platforms takes a great deal of manual effort. The agency wanted to coordinate it without a dedicated full-time person.

Challenges

Several brands, one system

Each business has its own tone, topics and audience, and content that suits one brand can damage another. The system had to keep each brand's profile separate and apply the right one every time it generates, so posts never blur together.

Three platforms with different rules

Facebook, Instagram and LinkedIn each expect different formats and offer different publishing routes. The build had to generate platform-specific variations and send each through the appropriate route, rather than posting one identical piece everywhere.

A platform limit on long-form articles

LinkedIn's API does not allow third-party tools to publish long-form articles. We could not automate that final step, so the system had to draft the article well enough that a person can paste and publish it quickly.

How the system works

Brand profiles

Each business has a stored profile in a database covering tone of voice, topics, target audience and platform preferences.

Scheduled generation

On a recurring schedule, the automation pulls the relevant brand profile and sends it to an AI model to produce platform-specific content variations.

Optional approval step

Content can go through a lightweight human approval before publishing, or run fully hands-off, depending on how much risk the business will accept.

Platform-specific publishing

Facebook and Instagram posts use official publishing tools. LinkedIn short posts and company page content go through a dedicated LinkedIn automation layer.

Long-form LinkedIn articles

LinkedIn's API does not allow third-party publishing of long-form articles. The AI drafts the article and a person pastes and publishes it manually, which is quick.

Design decisions

Brand profiles stored in a database

Tone of voice, topics, audience and platform preferences live in stored profiles that the generation step reads each time. Adding or adjusting a brand means editing a profile, not rebuilding a workflow, which keeps the system manageable as the agency grows.

Approval that can be switched on or off

Content can pass through a lightweight human approval before publishing, or run fully hands-off. We made this configurable because how much risk a business accepts on public posts differs from one client to the next.

Human publishing for long-form articles

Rather than looking for a workaround to LinkedIn's limit, we accepted it. The AI drafts the article and a person publishes it, which is quick and keeps the system inside the platform's rules.

Result in production

One system manages content for multiple businesses across three platforms. Automation covers the repetitive work, and people focus on approvals and long-form publishing.

One system covers content for several businesses across three platforms.

Automation handles the repetitive work of drafting and posting, so the agency does not need a dedicated full-time person for it.

People spend their time only on the parts that need them, namely approvals and long-form publishing.

What we would do differently

We would formalise the internal operations as small SaaS-style surfaces earlier, rather than composing n8n and Airtable. It ages better as the agency grows its headcount.

Who this pattern fits

Teams with similar volume, compliance or coordination challenges can use the same approach. The stack is stable, self-hosted where it matters and designed for a small team to operate.

Tools used

n8n

Workflow runtime and orchestration.

Claude API

Reasoning and structured outputs for content generation.

Supabase

Database, authentication and storage, including brand profiles.

Zernio

Publishing integration in the production stack.

Unipile

The integration layer used for LinkedIn.

Related

  • Custom AI Systems
  • Social Media Content Automation

Frequently asked questions

Does every post need human approval?

Not necessarily. Content can go through a lightweight human approval step before publishing, or run fully hands-off, depending on how much risk the business is willing to accept. The choice can differ from one brand to another.

How does the system keep each brand's voice separate?

Each business has a stored profile covering tone of voice, topics, target audience and platform preferences. On a recurring schedule, the relevant profile is sent to an AI model, which produces platform-specific variations for that brand.

Can it post long-form articles to LinkedIn?

Not directly. LinkedIn's API does not support publishing long-form articles, so the system drafts them and a person publishes them manually, which takes minutes rather than hours.

Services · Use cases · Industries · Case studies · Products · Free AI consultation · Book an AI consultation · AI readiness checklist · Automation ROI calculator · Live demos · About · Contact · Privacy Policy · Terms of Service

© Erudience Ltd · Contact · contact@erudience.com · +44 7848 146483

ERUDIENCE