Build-in-public, not a highlight reel

What's actually working. What isn't. Written down.

Two threads: monthly honest reviews from the live IT support business, and weekly notes mined from a real, shipped AI-assisted build. No "I make £1,000 a week" posts — the struggle is the credibility.

Open call — one slot

I'm looking for one sceptical local business to work with — free

A real Working Diagnosis, done properly, at no cost, for one business — in exchange for the honest, public story of how it went. Read the full offer →

Start here

What is AI, actually? A plain answer, not a sales pitch

What ChatGPT/Claude/Gemini actually do, why agents beat chatbots, whether it's hype, and who's really qualified to advise you on it in 2026 — read this first if any of it feels like a mystery.

Foundations
Try this

Five AI prompts to try today — no experience needed

Free, copy-paste prompts — check a marketing email, spot buying signals in customer notes, clean up messy phone numbers, and more. Five minutes, no sign-up.

Foundations
Philosophy

GIGO: the 1981 lesson that still runs my AI consultancy

Garbage in, garbage out — a rule learned coding on a ZX81 in 1981, and the whole reason OptimisedLead exists. The career evidence behind the philosophy.

Philosophy
Track B — business owner

One month on: what actually moved for Markus IT Support

The monthly honest review — ad spend that failed, the channel that's showing early signs of life, and what changed this month.

Monthly review
Track B — what didn't work

A chatbot is a glorified search bar — I built one to find that out

I built a chatbot for Markus IT Support. Users found it condescending, tried to break it, and trusted the business less. Here's why I pulled it.

Build-in-public lesson
Track A — build log

What building a bespoke platform in six weeks with AI actually looked like

Mined from a real, shipped build history. Outcome and process level only — no names, no code.

Build log
Craft — AI spend

How you waste money on tokens (and how routing fixes it)

The practical, unglamorous craft of managing AI cost — cheap models for grunt work, premium models for decisions that matter.

Craft note
Craft — how it works

Inside Hermes: one office, several specialist models

What multi-model routing actually is, and how Hermes — the system I built myself — routes work across AI models with a human always signing off.

Craft note
More on the way

This hub grows weekly (mined build-log content) and monthly (the honest business review). Check back, or come back once you've read a piece that speaks to your own bottleneck.

Send a quick note