AI Systems Architecture Blog — Agentic AI, RAG Pipelines, Multi-Agent Orchestration
On enterprise AI implementation — agentic AI deployment, RAG pipelines, multi-agent orchestration, and what it actually takes to ship production AI systems at enterprise scale.
Nobody Asked It To Is Not a Defense
It's a true sentence that functions as an admission. The fix is not a human approving every action — it's four fields, filled in once per class of action.
Authority Is Granted in Four Places
When a team goes looking for how their agent got permission, they open the identity system. That is one surface of four — and the only one with an owner, a review cycle, and an auditor.
Your Agent Is Wearing Your Badge
Permission systems were built around people who eventually leave. Agents never do. The credential you cloned in March is still live, still fully scoped, and still attached to whoever it was copied from.
The System Did It
Your logs record what your agents did. They do not record who agreed to it. Access and authorization used to be the same event. Agents separated them.
Trust Is a Checklist, Not a Feeling
A few weeks ago, my own system tried to hand me a fake — fluent, plausible, and wrong. A checklist question caught what inspection never would have. Here's the pre-flight for any agent with persistent memory, before it touches production.
Where Governance Can Actually Reach
Agent memory isn't one thing — it's four distinct surfaces, each carrying different privileges, each failing differently, each needing its control in a different place. Here's the map: every door, and where the lock goes.
The Prompt Is No Longer Where the Behavior Lives
On July 28, the protocol connecting AI agents to their tools stopped tracking who was talking — on purpose. It wasn't an isolated decision. Every layer of the stack is making the same move, and your audit surface hasn't followed.
Your Agents Remember Things Nobody Reviewed
Your system prompt went through security review. Your prompt library is versioned. The memory store now driving your agents' behavior went through neither — and your name is on the output either way.
The Two Failure Modes Nobody Formalizes
Every AI team can tell you what happens when their system works. Ask what happens when it fails silently, and the room goes quiet. Two disciplines close that gap — and almost no one has named them.
Your Corpus Is Lying to Your Model
When an AI system gives a confidently wrong answer, everyone audits the model. Almost no one audits what you fed it. The failure is usually in the corpus — and corpus quality is measurable.