Latest long read
Every question your AI readiness review asks was answered months ago
The last six days of a sixty-day series, and the pattern is that operability gets bought early or it does not get bought at all.
Maps, decisions and failures from enterprise systems
Real systems,
revealed.
How enterprise systems fit together, as maps you can open: what each part does, what I decided and what broke. AI first, because that is where I build now, on the data, integration, cloud and security platforms it depends on.
entries124
components mapped50
series published18
last entryOct 17, 2026
The whole estate
Six maps, one company
The AI application depends on four platforms, and security and governance apply to all five. So far two maps are interactive, one is written up and three are outlines.
Used byAgentObservability
Used byEdgeLLM GatewayModel
Written up · 4 components
Security & Governance
Identity, access and compliance, across all five layers
Start here
Where should I start?
- New to AI systemsYou have been asked to add AI to a product and want the ground truth first.
- Already shipping RAGRetrieval works in the demo; citations, freshness and access control do not yet.
- Putting agents in front of usersYou need permissions, budgets and stop conditions that hold up.
- Architect reviewing a designYou own the architecture review and want the whole system on one page.
- Owning evaluation or operationsYou decide what blocks a release and what wakes someone up.
- Building the platformTools, gateways and integrations that many AI apps will share.
- Hiring for AI systems workYou want to see how I design and explain systems before a conversation.
Read in order, or by question
Pick a track or a lens
8 five-day tracks of five short parts each, from the first decision to running it in production. Or follow a path: Agents in production, without a shared robot, Enterprise RAG, from problem to production, New to RAG, in order.
Tracks
Lenses: the question you actually have
- AI & Models63 entries
What does the model actually do, and where does it need the system to help?
On the map - Data & Databases45 entries
Where does data live, how does it move and how fresh is it?
On the map - Networking5 entries
Which paths does traffic take, and where does it leave your network?
On the map - Security30 entries
Where are the trust boundaries, and what can cross them?
On the map - Identity & Access7 entries
Who is acting, on whose behalf, with which permissions?
On the map - Integration24 entries
How does this connect to the systems the business already runs?
On the map - Compute & Infra21 entries
What runs where, and how does it scale and fail?
On the map - Observability & Ops22 entries
How do you know it is working, and what do you do when it is not?
On the map - Cost & Governance14 entries
What does it cost, who approved it, and what is on the record?
On the map
Recently written
The journal
We chose Azure AI Search for retrieval because ingestion and filtering stay simple
Why our retrieval runs on Azure AI Search: our Databricks ingestion pipelines feed it directly, filtering is simple, and several semantic configurations share one index.
We built our own TypeScript orchestrator, and stayed on it
Why our enterprise AI platform runs agents on a TypeScript execution layer we started almost two years ago, and why reaching near-parity with the frameworks argued for staying.
Sensitive data is cleaned at ingestion, with Presidio, before anything is indexed
Why our ingestion pipelines run Presidio over sensitive documents before chunking and embedding, and what that choice commits a team to: tuned detection, masking that is final for the index, and one more step for every document.