The working library / 87 articles
AI & systems
Practical workflows, clear boundaries and the work of turning an AI prototype into a dependable tool.
Field Notes explore an idea. System Maps lay out a process. Build Logs show the work behind it.
Browse topics / AI & systems
1–11 of 11 articles
Build Logs · AI & systems · 3 min read
Inside the content engine: source, draft, review.
A look at the structure of the content engine and why saved material, generated drafts and approved rewrites are separate.
Build Logs · AI & systems · 2 min read
Turning a script library into a useful journal.
How the content library became articles, topic collections and a source record rather than a bulk transcript dump.
System Maps · AI & systems · 2 min read
Measure AI cost per useful result.
Map context, retries, tool calls and review time before optimising a model bill.
Field Notes · AI & systems · 2 min read
The prototype is the beginning of the cost model.
A quick AI-assisted demo can test an idea while leaving most production obligations unresolved.
System Maps · AI & systems · 2 min read
Automate the preparation, then review the prospect.
A lead workflow that separates discovery, evidence, qualification and a deliberate contact decision.
System Maps · AI & systems · 2 min read
Read the skill before giving it a tool.
Review an agent extension for purpose, permissions, provenance and observable behaviour.
System Maps · AI & systems · 2 min read
The exception path is part of the AI workflow.
Define what happens when an agent is unsure, blocked or about to exceed its authority.
System Maps · AI & systems · 2 min read
Give the agent access that matches the job.
Map data, tools and consequences before connecting an assistant to business systems.
Field Notes · AI & systems · 2 min read
You can start without knowing every implementation detail.
Use a clear problem, a small scope and observable acceptance criteria when building with AI.
System Maps · AI & systems · 2 min read
Map the responsibilities before choosing the stack.
Separate interface, permissions, background work, retrieval and model calls in an AI application.
Field Notes · AI & systems · 2 min read
Adopt AI around a task, not a historical prediction.
Use the pace of technological change as a reason to learn, not as evidence that every new tool belongs in the business.
Follow a question
Explore the collections.
Built from my working notes, scripts and documented projects, with AI assistance in adapting and organising the library. Research and published case studies are linked in the articles.
How this library was built RSS feed