Writing
Plain-English essays on how AI systems actually work; agents, connectors, skills; and what they change for people with real work to do.
- 3 min read
Your best model should be the manager, not the worker
Orchestration explained without jargon. One expensive model plans and reviews, cheaper ones do the legwork in parallel, and the work comes back faster than doing it yourself.
- 3 min read
Why Claude says it can't see your Shopify store
A real story about a connector, and what "connecting AI to your systems" actually means, including why prompting harder never fixes it.
- 4 min read
What is an AI workflow? From chat to repeatable work
Most people use AI one question at a time. Chaining the steps, so each output becomes the next input, is the shift that matters; it takes the carrying off your desk and leaves the deciding with you.
- 3 min read
What is an AI token, and why did you hit your usage limit?
Usage limits are measured in tokens, not questions. A plain explanation of what a token is, what quietly burns through them, and how to stop hitting the wall.
- 3 min read
Intelligence is an API now
Most knowledge work is the same shape. Take information in, apply judgement, produce something. The judgement part just became something you can call on demand.
- 2 min read
How much thinking does this actually need?
Reasoning modes make models slower, pricier, and sometimes worse. When to turn the thinking up, when to turn it down, and why overthinking is a real failure mode.
- 3 min read
What is an AI agent, actually?
What the word "agent" means, explained with an ordinary office task.
Put it to work
Try the ideas on a real task.
The £49 AI for Marketing book explains the ideas in plain language, with familiar examples and the reasoning behind each answer.
Email notes
Practical AI, in your inbox.
Occasional notes from Abai on marketing workflows, useful examples, and ways to work together.