Articles

AI integration insights

Articles and implementation notes on AI operating models, workflow systems, governance, and the practical work of making AI useful in daily execution.

Latest thinking

AI integration insights

Notes on operating models, implementation patterns, workflow governance, and the practical work of making AI useful inside real organizations.

A laptop displaying an analytics dashboard beside a printed sheet and a potted plant.
Latest articleNov 16, 2026

Browser agent, deterministic script or direct API? Choose by the action, not the demo

Choose the least complex dependable method for each web action by comparing direct APIs, deterministic scripts, browser infrastructure and agents.

A hand moves one wooden cube beside a structure assembled from matching cubes.
ArticleNov 11, 2026

Bolt vs. Lovable vs. Reviewed Development: Test Portability Before Choosing an AI Builder

Compare Bolt, Lovable and reviewed development by testing whether a real internal application can be reconstructed and operated outside its original builder.

A blank white envelope lies on a wooden desk in front of an open laptop and a holder filled with pens and scissors.
ArticleNov 10, 2026

Retries Are Not Recovery: Designing Reconciliation for AI Workflow Side Effects

A practical decision guide for handling uncertain external effects with idempotency, verification, compensation and owned reconciliation.

Open metal filing-cabinet drawer filled with rows of paper index cards and labeled dividers.
ArticleNov 9, 2026

Design Source Permissions, Freshness and Deletion Before Building a Company RAG System

A pre-build framework for deciding source authority, preserving permissions, propagating deletions and testing whether a company RAG architecture is safe to approve.

Three coworkers plan around a laptop displaying a calendar while one person writes in a notebook.
ArticleNov 6, 2026

What Is AI Integration? How It Works in a Business Workflow

AI integration connects a model to company data, systems, decisions and operating controls so it can contribute to a real business workflow.

Two colleagues discuss a branching flowchart and interface sketches drawn on a whiteboard.
ArticleNov 5, 2026

Durable execution or a job queue: what an AI workflow actually needs

Separate workflow recovery, external outcomes and business-record reconciliation to choose between durable execution and a simpler queue-backed design.

A headset-wearing office worker reviews a printed form on a clipboard.
ArticleOct 29, 2026

Before a Voice Agent Changes an Appointment, Claim, or Payment: Define the Protected Action

Define the first consequential action in a voice workflow, then set its authorization, commitment, transfer and recovery rules before granting live access.

A person compares a printed financial document with an online form on a laptop.
ArticleOct 28, 2026

Do You Need an Agent Framework? Direct Code vs. LangChain, LangGraph, and Microsoft Agent Framework

Choose between direct application code, LangChain, LangGraph and Microsoft Agent Framework by comparing control, state, approvals, side effects, recovery and maintenance.

A thick stack of documents marked with dozens of colorful index tabs.
ArticleOct 27, 2026

RAG Has Two Quality Gates: Finding the Right Evidence and Producing a Defensible Answer

Evaluate RAG as an evidence supply chain with separate acceptance checks for retrieval quality and answer support before approving a pilot.

Rows of drive bays in a server rack with green status lights.
ArticleOct 14, 2026

How to Run a Recoverable n8n AI Workflow with Queue Mode, Error Routing, and Safe PostgreSQL Writes

Configure queue workers, retained execution evidence, error routing, idempotent PostgreSQL writes, verification and controlled recovery for an n8n AI workflow.

Close-up of a dark computer console showing repeated red “Failed to load resource” error messages.
ArticleOct 10, 2026

How to set up an n8n error workflow for failed AI automations

Configure an n8n error workflow that sends redacted, minimal-data failure alerts, preserves useful investigation context in n8n and can be verified safely.

Abstract workflow planning on a whiteboard
ArticleFeb 17, 2026

Designing AI operations without noise

How to make AI part of the operating model without creating another disconnected layer of tools, dashboards, and coordination work.

Technical implementation work at a terminal
ArticleFeb 5, 2026

Where AI implementation usually breaks

The recurring failure pattern is rarely the model alone. It is usually weak workflow ownership, missing context, incomplete integrations, and absent production controls.

Team reviewing operational work from above
ArticleJan 27, 2026

Monitoring the AI operating layer

Once AI becomes part of daily execution, the work shifts from launch to keeping the system observable, accountable, reliable, and useful as conditions change.

Subscribe

Stay close to the work.

Receive selected updates on AI operations, implementation patterns, governance, and new offerings.