Not sure where AI agents could help your team? Take the 5-minute assessment
Field notes

Notes from the work.

Methodology essays from inside live engagements. Architecture rules, governance patterns, and the failure modes we have designed around. Three to start.

Most consulting-firm "thought leadership" is written to attract leads, not to communicate substance. The result is broadly correct, locally vague, and mostly forgettable.

This section is the opposite by intent. The essays are written for one reader: the senior operator who needs the methodology to work in their specific context. The one who wants the load-bearing detail, not the surface frame.

Each essay names a pattern, explains why we use it, and explains what breaks when it is missing. The patterns come from real engagements. The essays are short by design (1,500 words or under).


Three essays to start

The current set.

01 02 03
Field note · Architecture · ~1,500 words

The three-phase pipeline pattern for audit-grade automation

Consolidation, structured extraction, and quality control as separate layers. Why mixing them produces unreliable systems.

Read
work deterministic judgment
Field note · Architecture · ~1,200 words

Deterministic where possible, judgment only where required

The architecture rule that decides what an AI should do and what code should do.

Read
RULE-001 RULE-007 RULE-014 flag flag ok
Field note · Governance · ~1,300 words

Flags only, never approvals

A governance pattern for AI in finance and compliance. The system surfaces. Humans decide.

Read

In draft

Essays scheduled for the coming months:

  • The two-scenario estimate. Why every effort estimate ships with Scenario A, Scenario B, and the named dependency between them.
  • Audience-filtered reporting without fact divergence. How the same status report goes to three audiences without becoming three different reports.
  • The production gap list. What separates a prototype from a production deployment, and why naming the gap is what makes the prototype useful.
  • Cross-industry transfer fallacy. When borrowing a model from Industry A to Industry B actually works, and the five questions to ask before assuming it will.

These publish on a roughly monthly cadence. No marketing automation. No email lists. No gated PDFs. The essays live on the site.

Want the methodology applied to your situation?

Reading the essays is the long way to learn whether this is a fit. A 30-minute call is the short way.