01OpenAI Presence: AI Agents Reach Production
OpenAI Presence shows where enterprise AI is heading: bounded jobs, governed data access, continuous evaluation, approved actions, and deliberate human escalation.
Read field note ↗Field notes · 2026
Ideas for systems that think—and still answer to people.
Practical perspectives on agentic AI, SQL, RPA, enterprise data, and the operating discipline behind useful automation.
01
02The Agentic AI Trust Gap Is a Data Problem
Enterprise AI agents are moving beyond pilots faster than organizations can govern the data, permissions, and decisions beneath them. Analytics teams can close that gap.
Read field note ↗
03Agentic AI Is Becoming an Operating System for Work
The real opportunity is not a smarter chat window. It is a coordinated layer that can reason, use tools, hand work off, and stay accountable.
Read field note ↗
04Why Stateless MCP Changes Enterprise AI
Scalable agent systems need context without fragile, server-bound sessions. MCP’s direction makes that architecture easier to reason about.
Read field note ↗
05Grounded RAG Agents for Enterprise Knowledge
Retrieval is not just a search feature. It is the evidence layer that keeps an agent useful, current, and honest.
Read field note ↗
06SQL Server 2025 Vector Search: What Changes
Vector search moves semantic retrieval closer to relational data—and gives SQL teams a practical role in the AI stack.
Read field note ↗
07Human-in-the-Loop Is the Control Plane
Human review should not be an emergency brake. It should be a designed layer for judgment, authorization, and learning.
Read field note ↗
08RPA and AI Agents: Better Together
RPA brings deterministic execution. Agents bring interpretation and adaptive planning. The combination is stronger than either alone.
Read field note ↗
09How to Evaluate an Agent Before It Reaches Production
A demo proves possibility. An evaluation program proves whether the system is safe, useful, and ready to operate.
Read field note ↗
10Prompt Injection Is a Permissions Problem
Better prompts help, but the durable defense is limiting what an agent can see, do, and authorize.
Read field note ↗
11SQL Performance Still Matters in the AI Stack
A brilliant model cannot rescue slow joins, stale pipelines, or unclear business logic. AI experiences inherit the quality of their data foundations.
Read field note ↗
12Building the Business Case for Agentic AI
The strongest case connects one constrained workflow to measurable capacity, quality, speed, and risk.
Read field note ↗Ideas are more useful in motion
Have a system, data, or AI problem worth unpacking?
Book time with Chris for a practical conversation, or request the résumé for a closer look at the experience behind the work.