The classic agency model is built on an idea that worked for decades: many people pooling their knowledge and skills to produce results together. This pooling was the value - and honestly the greatest joy of our profession. You needed the team to make complexity manageable.
With autonomous agents, this logic is beginning to dissolve. A well-connected agent can already handle research, copy, image, publishing, and iteration in a single run - if it knows the brand, sources, and tools.
What we experienced ourselves
We started with the expectation of primarily speeding up through AI. Then we tested it in practice - and the results surprised us more than we would have liked.
Our internal agent processes conversations from meetings, derives topics from them, formulates posts, creates them in the system, and generates matching visuals directly from brand context. What we did not plan: it began to derive missing connections and set them up itself. We gave it access, goal, and freedom - it organized the rest on its own.
This is no longer a copilot. This is an executing instance that carves out its own workspace.
The tension between ambition and business model
Agencies are supposed to guide brands strategically while simultaneously functioning as production capacity. Fees are based on hours and team sizes rather than impact. The pressure to secure projects and keep resources utilized remains high.
What counts now
When autonomous agents take over operational execution, the bundling of many roles loses economic justification. Then what counts is not the size of the team, but: how clearly the brand is defined, how accessible knowledge is stored, and how openly systems communicate with each other.
For us at AN, this insight was the occasion to expand our brand agency with Open Wonder - an AI-based brand infrastructure that meets exactly these new requirements. Not as a replacement for creative work, but as the foundation for a new way of organizing it.