Founding engagement · Editorial knowledge and AI workflow pilot
Keep the knowledge. Improve the next decision.
Teams collect research, interview customers, edit stories and make judgement calls every week. Then the project ends and much of the learning becomes difficult to recover. The next brief starts without the questions, evidence gaps and lessons that could have made it better.
I help a small B2B editorial or marketing team test one better way of working: a dependable evidence and decision trail, supported by AI where it is useful and subject to human verification throughout. This is a focused founding pilot, not a promise to transform an organisation's entire knowledge infrastructure.
The use cases
Choose one expensive point of repetition.
That might be a customer-story process where similar interviews are commissioned without consulting previous gaps, a research programme whose evidence is hard to reuse, or an editorial approval cycle where the same issues recur because no one has retained the reasoning.
The workflow does not have to use GitHub, build a new website or replace an existing CMS. The right starting point is the systems and permissions the team can actually work with.
What a pilot would involve
Map the current workflow
Identify where evidence arrives, where decisions are made, what gets lost and who approves what. Agree a bounded set of materials suitable for testing.
Build the evidence and judgement layer
Prototype an evidence register, source references, question and gap capture, retrieval rules or a repeatable editorial hand-off. Use AI only where it can improve a specific step.
Test it against real work
Run a small comparison: can people locate the source, recover the previous decision, spot unsupported claims and prepare the next brief more reliably?
Decide whether it deserves to scale
Record what worked, what failed, the time and review burden, and what the team would need to own. Retain a usable workflow specification even if the prototype should go no further.
Deliverables and boundaries
A limited test with a useful result.
The proposed output is a workflow map, one working evidence or retrieval prototype, a small evaluation record, and hand-off documentation covering ownership, source authority, review and limitations. Scope and duration are agreed before work begins; a three-to-six-week pilot is an indicative starting point, not a fixed promise.
No confidential customer or employer evidence should be uploaded without the right permission and an agreed handling process. AI suggestions are not treated as verified facts. The team retains responsibility for acceptance, approval and publication.
What has been demonstrated
The method is working in public. Multi-team adoption still needs testing.
Editorial Intelligence itself uses separate stores for method and publication, structured content relationships and explicit rules for retrieval, revision and human review. That is a functioning independent example, described in the publishing-system implementation guide.
My professional editorial work has also exposed repeated issues in customer-story capture, approvals and knowledge reuse. Those observations help define what ought to be tested; they are not proof that a client team has adopted the proposed workflow or achieved a quantified return. A founding engagement would produce the first independent adoption evidence.
Start with one repeated problem.
Tell me which editorial decision keeps being remade, what material the team already has and who needs to use the result. We can work out whether a bounded pilot is worthwhile before discussing tools or a larger build.