A brand system becomes operational when AI can use the right part at the right time
This anonymised case study shows how a global professional services organisation moved from a brand website designed for human browsing to a structured operating layer organised into fifteen AI skill sets. The source guidance covered far more than tone and visual identity. It included colour, typography, photography, video, co-branding, recruitment, and other context-specific requirements. Loading the whole site into every AI task would create noise, increase the chance of conflicting guidance, and make decisions difficult to explain.
We structured the full guidance system into themed capabilities that can be selected according to the task. The resulting skill sets allow an agent to retrieve the appropriate rules for a policy question, a creative brief, a draft, or a finished-work review without treating the brand site as one undifferentiated document library.
The challenge
Brand guidance had grown into a rich digital system, but AI access was still page-based. A model could retrieve text from the site, yet it did not reliably understand:
- which guidance family governed a particular task;
- how rules related to conditions, examples, and exceptions;
- which information was authoritative;
- when several guidance families needed to be combined; or
- why a recommendation had been made.
Some of the most important meaning also lived in images and designed examples rather than page text. A scrape could capture words while missing the rule those examples communicated.
Structuring the guidance
We decomposed the site into governed rules, relationships, examples, conditions, and source references. The content was then organised into fifteen themed skill sets, including areas such as colour, typography, co-branding, recruitment, photography, video, and verbal identity.
Each skill set has a defined purpose and knowledge boundary. That makes it possible to load the guidance needed for the task rather than flooding the model with every available rule.
The structured outputs were prepared in machine-readable forms, including Markdown and JSON, so they could support direct retrieval, API delivery, evaluation, and future agent workflows.
Three cross-map operating skills
In addition to the themed guidance sets, we designed three skills that work across the map:
- a policy skill for answering questions against approved guidance;
- a creation skill for applying relevant rules while producing a draft; and
- a review skill for checking completed work and explaining any correction.
These skills provide an operating layer above the source knowledge. They determine how the guidance is used, while the underlying map retains the rule, condition, relationship, and provenance behind every response.
Why task-specific loading matters
A recruitment asset should not require the same guidance bundle as a photography brief or co-branded proposal. Task-specific skills reduce irrelevant context and make the applied rule set easier to inspect.
This also improves maintainability. A change to typography or recruitment guidance can be reviewed and released in its own governed area without rebuilding a single monolithic prompt containing the whole brand system.
Outcome
The organisation now has a structured representation of its full brand site across fifteen themed skill sets, ready for brand-team review. The system creates a practical foundation for governed policy answers, content generation, finished-work review, and future asset evaluations.
The work remains subject to approval by the organisation’s brand team. That distinction is important: structuring makes the guidance operational, but accountable owners still decide when each skill and rule set is approved for production use.
What this case study demonstrates
An enterprise brand cannot be reduced to one prompt or one retrieval index. A skill-based model keeps the guidance modular, traceable, and relevant to the work being performed, allowing AI capability to grow without weakening brand ownership.