Text
Copy for a channel, a market and a language. From approved building blocks, not from a model's memory.
Platform
The core in depth.
Two layers, stacked. Apps, your own agents and partner tools dock on top. That order is the point: the core first, the features after.
01Brand core
The brand, decided once and closed.
- Every element is decided, approved and carries its source.
- Every element carries a knowledge state: known, unknown, deliberately undefined, not applicable.
- Known-false values are stored as false, with a pointer to the correct one.
- Gaps stay gaps. Nothing is inferred just to have an answer.
02Control layer
The layer that decides what gets through.
- Guidance from the standard, guides and rails per job.
- Context is narrowed: market, channel, language, brand.
- The result is a decision: pass, warn, block, or route to approval.
- Scored by a council of models, not by a single one.
03Connectors
The core does not work alone.
- DAM and MAM for assets with origin and validity.
- Email and campaign systems for delivery.
- Your own model over MCP or CLI.
- Product and master data as the source for technical values.
04Native apps
What runs on the core today. Every app reads from the same truth and leaves the same decision record.
Brand check
Text, image and PDF checked against the brand. With a reason and a source, not with a score.
Image check
Colour, logo, imagery, format. Whatever is not supported gets flagged.
Translation
Multilingual, using the brand glossary. Terms that are fixed stay fixed.
Print templates
Templates with checked colour values and placements, CMYK included.
Digital assets
Assets with origin, validity and approval state instead of a folder full of file names.
Brand knowledge
Questions to the brand. The answer names its source, or says there is none.