I’m Andrew Bembridge, Head of Digital Infrastructure and Automations at a Canadian tech company.

My career has been about following useful signals into the highest-value gaps and turning them into systems. Support, marketing, web, operations, and now the AI work that sits on top of all of it. I see businesses as a graph of handoffs, not a stack of departments. The interesting work tends to live in the handoffs nobody owns yet.

Most AI writing focuses on one piece: the API, the model, the agent. I write about the rest. The WooCommerce store, the WordPress site, the CRM, the Zap graph, the Cloudflare config, the autonomous engineering harness, and the glue that holds it all together. This is the working notebook.

What I work on

Agentic harness. I’m building Loopcycle, an autonomous orchestration platform. Engineering workflows that plan, execute, and review without me babysitting every step. What does “autonomous” actually look like when it’s running your overnight queue? I have opinions, and most of them were earned the hard way.

E-commerce and WooCommerce. Where AI earns revenue in a merch stack (product copy, support, recommendations, fraud, ops automation), versus where it is just a demo with a checkout button bolted on.

Customer support and CRMs. Where LLMs belong (enrichment, summarization, routing) and where they break things (anything that writes back to the source of truth without a human gate).

Zapier, Make, n8n. The new low-code. AI nodes inside glue. When a fourteen-step Zap with a GPT step beats writing a service, and when it absolutely does not.

WordPress. Yes, still WordPress. It’s quietly one of the best AI-augmented stacks in 2026, and most engineers are too embarrassed to notice. I’ll explain.

Production reliability and security. Retry logic, idempotency, audit trails, secure APIs, zero-trust rulesets, and the boring engineering that keeps AI agents reliable past the demo.

The generalist view. What transfers between areas. What doesn’t. The AI playbook from a generalist who is tired of the specialist takes.

How I write here

First person. Opinionated. Specific tools, named platforms, real failures. No “five trends to watch” posts. No hedged thought-leadership. If I haven’t actually run the thing, I don’t write about it.

Two registers:

  • Essays. Where I commit to a take. Roughly monthly.
  • Lab notes. Shorter, rougher write-ups of things I shipped or broke this week.

Loopcycle

Loopcycle is the orchestration platform I’m building. A node-based system where every workflow is visible, composable, executable, and testable. Think n8n, but purpose-built for AI orchestration. Most of what I learn about agentic systems comes from building it.

Work with me

I take on a small number of consulting engagements each year, usually with teams who want help wiring AI into systems they already own (commerce, CRM, support, content), or who are trying to get autonomous engineering past the demo stage. If that is you, reach out on X or LinkedIn.