Insights
Notes on AI automation that pays for itself.
Longer pieces on the decisions that are easy to get wrong: which workflow is worth automating, how to measure the hours, what a document pipeline needs, and where the AI Act becomes engineering.
What this section is for
Each piece answers a question an operations lead or an engineer actually asks, with the trade-offs and the numbers in the open. No thought-leadership, no listicles. If a post recommends an approach, it also says when that approach is wrong.
01 / All notes
Read in the order that matches your problem.
Start with the decision you are facing, not the newest post.
Where AI automation actually saves time (2026)
A practical model for choosing a workflow: volume, handling time, exception rate and the cost of the review step. Includes a one-page decision test.
EU AI Act Article 50: the engineering work due by 2 December 2026
Article 50 transparency obligations are due from 2 December 2026. What engineering teams must build, what is legal work, and how to avoid overclaiming.
When fine-tuning actually beats prompting (2026)
A decision guide for engineering leaders: the four problems fine-tuning solves, the volume threshold where it pays, and how to measure whether it worked.
How to evaluate a RAG system before your users do (2026)
A practical evaluation protocol for RAG: golden sets, retrieval metrics, rubrics, LLM judges and a CI gate that catches regressions before users do.
What on-premise LLM deployment actually requires (2026)
Hardware sizing, quantization, serving, security and operations: an honest checklist for running an LLM on infrastructure you own, without the demo math.
Next step
Want this applied to your workflow?
The posts describe the method. A fixed-scope audit applies it to your process, your exceptions and your tools.
Prefer email? [email protected]