Your AI Marketing Tools Are Only as Good as What You Document
Undocumented business knowledge produces generic AI output. Here's how I capture client context and what it does to the quality of AI-assisted campaign work over time.
What is happening
Anthropic published a breakdown of the internal AI workflows they have built for their own teams - hundreds of them, covering code review, deployments, diagnostics, and more. The document is aimed at developers. But buried in it is a finding that I think matters more for marketing teams than for any engineering organisation.
The finding: AI performs well when it has specific, documented knowledge to act on. Generic instructions produce generic output. The teams getting the best results from AI are the ones who have taken the time to write down what they know - in enough detail that the AI can actually use it.
Anthropic calls the important parts “gotchas sections.” The common failure points. The things that look like they should work one way but actually work another.
What I learned from this
Most marketing teams do not document anything. The knowledge of how the business works, which campaigns behave unusually, which audiences underperform in certain markets, which bid strategies need adjustment going into Q4 - all of that lives in the heads of the people who have been doing the work long enough to have made the mistakes.
I have seen what this looks like when you run AI against an undocumented marketing operation. The outputs are technically correct and contextually wrong. A generated brief that misses the brand nuance that took two years to develop. A keyword list that is correct by category but wrong for the specific audience this business serves. A reporting summary that draws the right conclusion from the data but does not know that this particular client excludes branded search from their ROAS calculation.
The AI is not failing. It does not have the knowledge it needs to succeed. The gap between what an experienced practitioner knows and what the AI has access to becomes visible immediately in the quality of the output.
This is not a new problem. Agencies and in-house teams have always had knowledge concentrated in a few people, transferred informally, and lost when those people move on. AI just makes the cost of undocumented knowledge visible in a way that was easier to ignore before.
What I recommend for your business
Start treating documentation as part of the work, not something that happens after the work is done.
Every time you make a decision that is specific to your business, your audience, or your campaigns, write it down somewhere that can be referenced later. Not in a formal document. In a running note that grows over time. The audience that performs well on Meta but fails on Google. The offer that converts in autumn but falls flat in spring. The landing page variation that looks worse but converts better.
That knowledge base becomes the context you give to AI tools when you use them. Instead of asking an AI to write ad headlines and getting generic output, you give it the business context, the audience knowledge, and the creative principles that have worked. The output improves substantially.
The teams that will do best with AI over the next few years are not the ones with the most advanced tools. They are the ones who have taken seriously the work of capturing what they know. The tools are becoming a commodity. The knowledge is not. Document it.