SANCHE.AI / TECH FIELDWORK
Tech that survives contact with the operating model.
A working notebook for people building at the Microsoft AI edge: Microsoft Foundry, Agent 365, identity, evaluation, data boundaries—and the operating choices behind them.
THE LENS
Not product news. Technical judgment for real operating environments.
Every note starts with a practical tension: what should an agent be allowed to do, which identity should it use, what evidence makes it trustworthy, and who owns it once the demo is over?
01 / MICROSOFT FOUNDRY
From prototype to a system you can inspect.
Models, agents, tools, traces and evaluations—translated into decisions on architecture, cost, data and ownership.
Official documentation ↗02 / AGENT 365
The control-plane question.
How agent inventory, identity, permissions, compliance and oversight change when agents become part of everyday work.
Official overview ↗03 / EXERCISES
One build. One boundary. One reflection.
Short, practical exercises designed to turn a technical feature into a deliberate capability—not merely another pilot.
How the exercises work ↓THE EXERCISE FORMAT
Build with a decision, not a feature list.
01 / Frame
One concrete work problem and the boundary an agent must respect.
02 / Build
A focused 30–45 minute exercise with a success signal and an evidence trail.
03 / Reflect
The operating implication: identity, ownership, governance, adoption or the next experiment.
Each future exercise will include a short reading marker—source documentation, a Frontier Firm lens and a question to apply in your own environment.
WEEKLY FIELD NOTES
The technical track stays connected to the leadership, operating-model and governance questions around it.
Sanche.ai 