Skip to content
A mini-series of systems

Before the code, there was a doodle.

Five working systems, each told the way it happened: the napkin sketch first, then what it grew into.

one made-up company, five systems 2:07 a.m. offer sent case 01 3:45 p.m. booking checked case 02 Oct 24 price drop caught case 03 Mon 9 a.m. trend spotted case 04 Tuesday story written case 05
Track one

Operations

The everyday work nobody notices until it goes wrong: the quote, the check, the price that quietly changed.

the idea: request in AI reads it + the playbook rules check it find options send offer …while everyone sleeps
Case 01

The 24-Hour Shift

A prospective customer writes comments in plain words on a quote request, at any hour. Minutes later, they have a real offer with real prices, checked against every partner’s rules.

the moment: “The AI missed the dog. The system caught it.”

Read case 01 →
the idea: the queue both papers read compare write it up keep score …every hour, whether anyone is watching or not
Case 02

Second Set of Eyes

Managers used to work through the whole checklist on every booking. Now, by the time they open one, most of it is already ticked, and only the calls that need their judgment are left.

the moment: Three lines match, two need fixing, and one is a manager’s call.

Read case 02 →
the idea: bookings till deadline today’s price? same thing? flag it person decides =? …every booking, until it can’t change anymore
Case 03

The Price Watcher

Prices drop after people book. This keeps an eye on every open booking, compares like with like, and speaks up only when the savings are real.

the moment: The $749 option looked cheaper. It wasn’t the same deal.

Read case 03 →
Track two

Creative

Finding what people care about, then telling those stories in a voice that’s unmistakably yours.

the idea: listen group score judge report to the writers …every Monday, before anyone’s had coffee
Case 04

The Signals

By the time a trend is obvious, it’s too late to lead with it. This system watches forums, videos, journalists and newsletters for the early signs, and each Monday tells the team which stories are likely to break out, and how likely.

the moment: It scored a story 1 of 5 and recommended it anyway.

Read case 04 →
the idea: idea list pick a writer research draft + check a person says yes ready …one piece at a time, in each writer’s own voice
Case 05

The Writers’ Room

It starts with a messy list of ideas, or even simply titles. The system assigns each idea to the right trained writer from a stable, does the research, validates the angles, and the AI writer drafts V1. Then a human editor joins the mix to iterate, until each piece sounds on brand and every claim is sourced.

the moment: “Research shows” was really “some evidence suggests.” Sent back.

Read case 05 →

The method behind all five.

Different jobs, same way of working.

  1. 1

    Find the bottleneck

    Where does work sit and wait? Where do the same mistakes keep slipping through?

  2. 2

    Write down the rules

    Each partner quirk and house rule goes into a playbook your people can read and fix as needed.

  3. 3

    Build the machine

    Hand the reading, checking and legwork to AI, with firm rules it can’t talk its way around.

  4. 4

    Keep a person in charge, and let it learn

    One switch, clear hand-offs, honest numbers. And every time someone corrects it, the system learns, so it gets better the more your team uses it.