AI marketing
needs operators.

I build AI systems for marketing work that keeps getting pushed to the backlog. This is where I write what shipped, what broke, and what I learned while building them.

Read the articlesGet the newsletter

What I write about

⚙️

AI systems

How I build agents that research, draft, test, route work, and stay inside review gates.

📈

Growth strategy

Retention loops, funnel math, pricing, and distribution from the operator side. The parts that decide whether AI turns into revenue.

🧠

Decision frameworks

How I decide where AI belongs, what to build first, and what to avoid when the demo is better than the operating reality.

Consulting|I run NorthSignal, where I help B2B SaaS teams design and build AI marketing systems. I handle strategy, architecture, and the hands-on build.
Visit NorthSignal →

Latest articles

View all →

The Two Questions I Use to Decide What to Automate Next

When someone asks me what they should automate first, I do not give them a tool list. I ask two questions that reveal the real bottleneck. Here is the framework, where it came from, and how I adapted it for AI agent work.

ai-automationai-agentsframeworksbuild-notes

I Run Three Claude Instances as a Team. It Costs $3.50 Per Feature and Cuts My Rework in Half

Instead of one Claude session trying to plan, build, and review in a single context window, I split the work across three specialized agents. Here is the architecture, the cost math, and the CLAUDE.md template that keeps them aligned.

claude-codeai-agentsmulti-agentbuild-notescontext-engineering

I Audited My Claude Code Setup and Found 90K Tokens of Context Waste

Every Claude Code session I was running burned 15 percent of the context window before I typed a single prompt. Here is the audit I ran, the three things that were burning tokens, and the cleanup that cut my agent costs by 40 percent.

claude-codeai-agentscontext-engineeringbuild-notes

Build Notes

One email every couple weeks. What I am building, what broke, and what actually worked. No generated filler.

Subscribing also gets you the AI Agent Eval Checklist the day it ships: a 4-layer framework for checking if your agents are doing good work or just doing work.