FIELD NOTES_
Home/Blog
Field notes
Working notes, not thought leadership.
Working notes, not thought leadership. What we have learned from shipping AI in regulated, complex environments. Audited numbers. Real failures. No SEO bait.
Field Notes
WHAT WE
LEARNED ON
THE WAY HERE.
Featured this week
STRATEGY
May 2, 2026 · 12 min
The Real Cost of an AI Pilot That Goes Nowhere
After auditing 240 stalled AI pilots across enterprise clients, we found three failure modes. None of them are technical. Here is what to do about each.
Read article
All field notes
THE ARCHIVE
IMPLEMENTATION
8 min
Build vs. Buy in 2026: A Decision Tree That Actually Works
Most build-vs-buy frameworks were written for SaaS. AI is different. Defensibility, data flywheels, and buyer-beware compliance change the math.
AK
Anya Kowalski
Apr 28, 2026
GOVERNANCE
10 min
EU AI Act, One Year In: What Compliance Actually Looks Like
A pragmatic compliance checklist from the trenches of three Tier-1 bank deployments.
PS
Priya Suresh
Apr 22, 2026
IMPLEMENTATION
6 min
Why Your RAG System Hallucinates and How to Diagnose It
Eight specific anti-patterns we keep finding in client RAG deployments, with concrete fixes.
SP
Sebastian Palacios
Apr 18, 2026
CASE NOTES
9 min
How We Cut Charting Time 41% Without Replacing Epic
The architectural pattern behind the NorthBay Health rollout and why we did not pull a rip-and-replace.
DK
Dr. Daniel Kim
Apr 14, 2026
STRATEGY
7 min
The 90-Day AI Audit We Run Before Any Engagement
Our discovery sprint, fully open-sourced. The questions, the scoring, and the readout template.
JR
Jordan Reyes
Apr 8, 2026
INFRA
11 min
Notes From Running 14 Production LLM Stacks
Cost, latency, and reliability numbers from a year of real deployments, with graphs.
NV
Nora Vetrov
Apr 2, 2026
GOVERNANCE
9 min
Bias Audits That Are Not Theatre
How to design a bias audit that produces decisions, not just dashboards.
PS
Priya Suresh
Mar 28, 2026
IMPLEMENTATION
8 min
Eval Frameworks: The Quiet Difference Between AI That Ships and AI That Doesn't
Eval is the most underinvested-in part of every client stack. Here is a starter kit.
AK
Anya Kowalski
Mar 22, 2026
CASE NOTES
10 min
The Architecture Behind a Real-Time Fraud Stack at <50ms
How the Kestrel Bank fraud-scoring engine works, end to end.
NV
Nora Vetrov
Mar 18, 2026
Field Notes weekly
ONE EMAIL.
FRIDAY MORNINGS.
A short note on what we are seeing in AI deployment, governance, and adoption. One useful idea per week, no fluff.
Subscribe
Let's build together
WANT US TO COVER SOMETHING SPECIFIC?
Reply to the newsletter or send us a topic. We turn 1 in 4 reader prompts into a full field note.
Start a project