One agent per brand to run eleven headless sites at GDA

GDA runs more than eleven corporate brands on a single headless platform in Magnolia. Each one has its own value proposition, tone and audience.

In a multisite environment the risk is not publishing too little: it is that one brand content shows up in another brand voice, and nobody notices until it is in production.

And at that scale, operating by hand means choosing between speed and governance.

• Eleven different value propositions: what works for one business unit is exactly what feels wrong in the next.
• Cross contamination: an operator with access to everything is an operator who can pick the wrong brand without the system stopping them.
• Sustained volume: landing pages, catalogs and campaigns that do not wait for someone to be available.
• No independent check: whoever produces the content cannot be the only one approving it.

Content published in the wrong tone looks just as polished as correct content.

We designed a multi-agent operating model where each brand has its own agent, and where isolation between brands is built into the architecture rather than left to the judgment of whoever is operating.

Per brand isolation and context partitioning

• A dedicated agent per business unit, with its own credentials, tokens, brand manual and editorial guidelines.
• No agent holds another brand material in context, so it cannot apply it by mistake.
• It is the same logic that governs per client isolation: the rejection has to happen before access, not after publication.

Autonomous content and landing page operation

• A pipeline for structured landing page generation on the headless platform.
• Catalog updates and copy adapted to each brand value proposition.
• Continuous synchronization with the content manager, so what the agent produces is what the editor sees.

Campaign activation

• Paid media tied to each brand launches and offers.
• Campaigns go live alongside the content that supports them, not separately.

Experimentation and cross curation

• Continuous multivariate A/B testing on copy and conversion formats.
• Cross curation between agents: whoever produces is not whoever approves, and the review comes from another role.
• Monitoring with performance alerts, so a drop is seen when it happens and not in next month report.

Architecture and technology

• Magnolia headless as the single presentation layer, decoupled from the activation channels.
• One agent per brand, with cross agent QA gates and strict context partitioning.

How generative AI is integrated and governed

• The solution integrates generative artificial intelligence services: language models that write, converse, extract and structure inside the workflow, rather than a rules engine with prewritten answers.
• Consumption is measured and billed in tokens, separating input, output and cache write and read, which are different line items with different prices and are accounted for separately.
• Telemetry attributes consumption per agent and per client, so the token budget is bounded and controlled per project instead of being read as a single invoice.
• Limits live in the record, not in the model: a record declares roles and caps, and which provider each role maps to is configuration that changes without touching the agent.
• The architecture is multi provider: the provider is fixed per agent and per session, never per request, and automatic retry against a different provider is deliberately switched off. If the provider fails, the agent stops and says so instead of silently continuing on another one.
• Context caching is the main cost lever, because rereading context accounts for most of the token volume in any agent operation. Sustaining the cache is what keeps the same task from costing several times more.
• And the minimum context window is declared per agent: a provider that does not cover it is ruled out by architecture, not by price.

Learn more about AI agents, Magnolia CMS and software development.

GDA moved from running eleven brands with a shared team to having eleven parallel operations that do not touch each other.

What is running

• More than 11 corporate brands operated in parallel on one headless platform.
• A dedicated agent per brand, with its own credentials, manual and editorial guidelines.
• Landing page generation and catalog updates with no queue.
• Paid media activated together with the content behind it.
• Continuous A/B experimentation with performance alerts.

What changes in governance

• Independence between brands stops depending on the operator remembering which one they are working on.
• Cross curation introduces a control that did not exist: someone other than the author reviews before publishing.

Why it matters

• At multisite scale, governance does not come from more procedure: it comes from separating contexts so the mistake is not available.