The journal
Notes from the plumbing.
How we work, written down. Field notes on designing, shipping and running AI automation inside real companies.
What to automate first.
Every company has a hundred candidate workflows. The first one you automate decides whether the programme survives its first quarter. Here is how we choose it.
What to automate first.
Every company has a hundred candidate workflows. The first one you automate decides whether the programme survives its first quarter. Here is how we choose it.
Shadow mode: shipping an agent without betting the company on it.
Before an agent is allowed to act, it runs alongside your team for a while and is scored against what they actually did. This is how that works.
Human approvals are a feature, not a fallback.
The best automation we have shipped still asks a person to click approve on the cases that matter. That is not a limitation of the technology. It is the design.
Integrations without rip-and-replace.
Your ERP is not going anywhere, and it should not have to. How we connect agents to the systems you already run, including the old ones.
Voice agents: what callers actually notice.
Callers forgive a slightly robotic voice. They do not forgive being made to repeat themselves. What matters when a machine answers the phone.
What an MCP server is, and why your company might want one.
A plain-language guide to the Model Context Protocol: what it does, what it does not do, and when exposing your systems through it is worth the effort.
Nothing matches that yet. Try another word.