About

The short version

Every business is a set of steps that turn a stranger into a paying customer. Software can now do a surprising number of those steps. Deciding which ones is the whole problem, and almost nobody has a method for it — they have enthusiasm, a demo, and a tool that was easy to buy.

Spine is that method. It maps how a business actually works, decides step by step what a machine may do and who stays answerable, and keeps a record showing the line is still being held after everyone has gone home.

Why it exists

Automating the wrong step is not a small mistake. The most common expensive one in the trades is generating replies to customer reviews: it is visible, it feels like progress, and it quietly damages the one thing every other improvement depends on — a customer who believes a person is listening.

A method that can tell you not to automate something is worth more than one that automates faster. That is the part of this that took the longest to get right, and the part that is hardest to copy.

Who is behind it

Christian Mandefro — a full-stack developer who builds AI systems and got interested in the gap between what these tools can technically do and what a business should actually let them do.

The method is built on established systemisation practice, credited in full where it is used. What is added is the decision layer: the classification, and the record that keeps it honest.

What is true today

The method is complete and written down. The software that supports it covers the first stage and the decision stage; the middle is deliberately unbuilt, because building tools for an engagement nobody has bought is a good way to waste a year.

No engagement has run end to end with a client. Nobody has paid for this. The first person to do so would be the first, and the terms should reflect that.

There is a version of this page that does not say that. It would be easy to write and it would be more persuasive for about a month.

How it gets tested on itself

The same diagnostic gets pointed at this project, on the same terms as it would be pointed at a client — including when the answer is unflattering. It has been. One instrument turned out not to distinguish between businesses well enough to be useful from the outside, and that is recorded rather than quietly dropped.

An instrument that flatters the person holding it is not an instrument.