What we saw across 500 AI tools in one enterprise
Watching the AI layer showed that tool count was only the surface problem. The larger issue was not knowing which work deserved to become a company default.
Field Note 01 · July 11, 2026
Resources for company AI teams.
Essays, field notes, and reports on AI work, spend, workflows, model economics, and company-owned AI.
20 resources shown
All resources
The same workflow showed up in seven departments
Seven teams had independently built versions of the same research workflow. Visibility turned that duplication into a reusable company asset.
Field Note 02 · July 11, 2026
Where token compression saved money and where it did not
Compression produced meaningful savings on repeated, context-heavy requests. It mattered less when prompts were already short or the model output dominated cost.
Field Note 03 · July 11, 2026
Oximy Joins Y Combinator W26: Building the System of Record for Enterprise AI
We're coming out of stealth and joining Y Combinator's Winter 2026 batch. Here's why we're building the visibility layer the enterprise has been waiting for.
Blog · 2026-01-27
You Can't Secure AI You Can't See: The Enterprise Visibility Gap
Enterprise AI adoption accelerated because it was useful. What didn't evolve at the same pace was visibility. Here's why that's the biggest risk to scaling AI safely.
Blog · 2026-01-23
The Case Against Instrumentation-Heavy AI Security
Why observability must survive developer reality. Security that assumes perfection does not scale - observability that survives imperfect behavior will endure.
Guide · 2026-01-18
When AI Goes Wrong, Who Is Accountable? The Ownership Gap Enterprises Haven't Solved Yet
As AI systems influence decisions, automate workflows, and shape outcomes, responsibility is quietly diffusing. When something goes wrong, the organization often discovers that no single team fully owns the system involved.
Blog · 2026-01-15
The Illusion of Control in SDK-First AI Security
Why Instrumentation-Based AI Governance Cannot Scale in the Enterprise. A critical analysis of why SDK-first approaches create false assurance as AI adoption expands.
Report · 2026-01-15
Security Starts Before the Model: Why AI Risk Emerges Upstream
Why AI risk emerges upstream of prompts, providers, and policies. The most effective security controls are those that shape decisions before they become irreversible.
Guide · 2026-01-08
AI Observability for Employee Activity: Why It's Different
Employee AI observability does not fail at the prototype stage. It fails when AI becomes a daily habit. Here's why tracking workforce AI usage requires a fundamentally different approach.
Blog · 2026-01-06
Why AI Security Starts Too Late
A Structural Analysis of Upstream vs. Downstream AI Risk. Understanding why model-centric security controls fail and how to shift security earlier in the AI interaction lifecycle.
Report · 2026-01-05
The Workforce Comes First: Why Most Enterprise AI Risk Starts With People
Why most enterprise AI risk starts with people, not agents. Organizations that start with workforce visibility gain understanding, trust, and control.
Guide · 2025-12-28
The Hidden Cost of Shadow AI
Why Employee AI Usage Is the Largest Unpriced Risk in the Enterprise. A deep analysis of Shadow AI as a distinct enterprise risk vector driven by human behavior.
Report · 2025-12-22
Build vs Buy AI Observability: What Enterprises Must Decide Early
Nearly every enterprise that reaches meaningful AI adoption encounters the same moment of clarity: how do we see what's happening without slowing everything down? That's when the build-versus-buy discussion begins.
Blog · 2025-12-20
The Fast Path to Safe AI Adoption: A 30-Day Rollout Plan for CISOs
A practical, thirty-day path from uncertain AI usage to defensible, governed adoption - designed for organizations that need to move now, not eventually.
Blog · 2025-12-12
AI Observability Is Not an Application Problem
Why Enterprise AI Governance Must Be Infrastructure-Level or It Will Fail. A comprehensive analysis of why application-centric observability cannot provide durable visibility at enterprise scale.
Report · 2025-12-10
AI Agents Change the Security Model: Why Observability and Control Matter More Than Ever
AI agents do not merely respond. They act. They browse, call APIs, write files, trigger workflows, and make decisions that have real operational consequences. This fundamentally changes the security model.
Blog · 2025-12-05
The Enterprise Guide to AI Observability at the Network Layer
Why durable AI visibility cannot depend on applications, SDKs, or developer behavior. A comprehensive guide to network-layer observability for AI systems.
Guide · 2025-12-05
From 'Acceptable Use' to Enforceable Policy: What Real AI Governance Looks Like
Most enterprises already have an AI policy. It sits alongside information-security guidelines and acceptable-use standards. On paper, it looks like progress. In practice, it rarely governs anything.
Blog · 2025-11-28
Shadow AI Is the New Shadow IT - Except It Moves Data, Not Just Apps
Shadow AI is not simply the next iteration of shadow IT. It is categorically more dangerous because AI systems move data, interpret intent, and generate outcomes in ways traditional tools never did.
Blog · 2025-11-22