Skip to content

Short answers and longer arguments

Writing

Most of these argue one thing from different sides: systems fail at integration and adoption, not at features.

0.1in-house

The people who should build your AI already work for you

Why internal AI capability beats an external consultant. Notes on building retrieval systems and internal platforms with the people who already know your operations.

The useful question is not what AI can replace. It is what it lets every person in a company reach that they could not reach before.
0.2process

Write the process before you buy the tool

Most software spend replaces thinking instead of supporting it. Notes on writing down the workflow before buying the tool, and what that saves.

A twelve-page magazine feature, an IDML pipeline in Node, and a contract repaired at XML level. What changes when a document stops being a file.
0.3systems

Static systems cost you in changes, not in licences

The real cost of vendor software is the waiting, not the licence. Notes on replacing static systems with platforms your own team can change.

How one person runs a production system for seventy people without it falling over. Process, not heroics.
0.4brand

Brand systems that outlive the people who built them

Brand architecture across several brands and legal entities, design systems that hold in production, and why company software needs a brand of its own.

The person who was not in the room is the only honest reviewer a brand system has.
0.5timing

The switch does not get cheaper if you wait

Postponing a platform or AI migration makes it more expensive, not less. Notes on sequencing the switch in a mid-sized company.

Waiting does not reduce the work. It increases the amount of history that has to move with you.