Brennan Brand skill
One of the first skills I built. The job it does is keep outputs from Copilot Cowork and Claude on-brand without forcing every document through a central gatekeeper.
The shift
The default behaviour of a general-purpose AI assistant when asked to produce a document, a deck or an HTML file is to produce something that looks fine and feels generic. Generic isn't good enough when the artefact carries a brand on it. The shift the skill makes is from generic output to output that uses Brennan colours, typography, layout conventions, shape language and the design system that sits behind all of it.
That sounds like a small thing. It isn't. The difference between an on-brand document and a generic-looking one is the difference between an artefact that can leave the building and one that needs reformatting before it can. At the scale AI assistants are being used inside the business now, that's hundreds of hours of reformatting work that just doesn't need to happen.
A representative example, with a fictional firm (Meridian Partners) standing in so the demo isn't tied to a specific real brand:
Why this was worth doing
AI assistants spread through the business faster than central guidance could land. People were producing real customer-facing output before the policy work had caught up. The choice in front of us wasn't "let people use AI or don't". It was "equip them to use AI well or watch the brand drift".
We didn't want every output to route through a central reviewer. That's a tax on adoption and a bottleneck for everyone involved. We did want everyone using AI to be able to produce on-brand output autonomously, with confidence, the first time. The brand skill is the mechanism. It runs in the background whenever someone asks Claude or Copilot Cowork to produce a Brennan-branded artefact, applies the right rules and produces output the team can stand behind.
What made it possible
The work the Marketing team did during the rebrand. The brand guidelines, the design system, the typography hierarchy, the colour palette logic and the layout conventions were all documented to a standard most internal brand programmes don't reach. That documentation is the input the skill encodes. Without it, the skill would have been a guess. With it, the skill is a reliable interpreter.
That's worth calling out because the bottleneck for skills like this isn't the AI side. It's whether the source material exists in a form that can be encoded. The cleaner the brand documentation, the cleaner the resulting skill.
What's next
The brand skill is one of a small family I'm building inside Brennan that make AI assistants behave like Brennan-aware employees rather than generic tools. Brand is the first because brand is the most visible. The same pattern applies to writing tone, document structures, sales process guardrails and Presales artefacts. Each one is a small intervention in how the assistant behaves by default. Together they're how an internal AI capability stops looking generic and starts looking like the company.