Understanding AI, from zero to the boardroom.
A series in order: what artificial intelligence is, how it acts, what it does in a company, what it costs and where the blind spots are. For people who run a business and do not write code.
Understanding AI ·
Nobody has explained to you what AI is. Let us start there.
First note in the “Understanding AI” series: what artificial intelligence is and is not, where it comes from, the three ways of building it and two sentences to explain it to anyone in your company.
Read the note ›Understanding AI ·
It does not think. It predicts. And that explains almost everything.
Second note in the series: what a language model does inside when you write to it, why its memory has a size, why it would rather guess than say “I don’t know” and what that changes in your company.
Read the note ›Understanding AI ·
Three different things. One word in the catalogue.
Third note in the series: what separates a program with rules, a robot that imitates clicks and a model that predicts; where each one fits; and the seven questions that take a sales presentation apart.
Read the note ›Understanding AI ·
Nine verbs, one jagged frontier.
Fourth note in the series: the real capabilities of a language model, where they fit in each department, what the serious studies have measured on productivity and why the same tool helps in one task and harms in the next.
Read the note ›Understanding AI ·
An agent is a loop. And loops are watched.
Fifth note in the series: the three pieces that turn a model into a useful system, the five patterns used to chain them, what separates an agent from a workflow and why a long chain of correct steps ends up failing.
Read the note ›Understanding AI ·
The project starts long before you choose a vendor.
Sixth note in the series: the nine steps for taking a process with AI from idea to production without it stalling as a pilot, with what the studies say about buying or building, about data quality and about what your team already does on its own.
Read the note ›Understanding AI ·
The model is the cheap part.
Seventh note in the series: the full anatomy of the cost, from the token to the hour of review, why the model is usually the smallest item, where the return comes from and how to do the sum with a step-by-step example.
Read the note ›Understanding AI ·
The pilot does not fail at the end. It fails in the first meeting.
Eighth note in the series: what the studies say about the failure rate of AI projects, the five underlying causes, the early symptoms an executive can see without knowing the technology and the way out of the eternal pilot trap.
Read the note ›Understanding AI ·
The law already applies. Since before AI.
Ninth note in the series: the three layers of law that affect a company using AI, data protection, the EU regulation and the Swiss approach, with the dates that really concern you, the cases that illustrate each rule and a list of what to do before asking your lawyer.
Read the note ›Understanding AI ·
A model obeys text. All of the text.
Tenth note in the series: what can go wrong in a system with a language model, from prompt injection to the invoice with no ceiling, with the three public cases that teach the most and a list of measures that do not require a security department.
Read the note ›Understanding AI ·
What nobody puts in the presentation.
Eleventh note in the series: what is ignored, omitted or considered irrelevant when AI is discussed, and which decides the outcome. Ten blind spots with their source, their symptom and their remedy.
Read the note ›Understanding AI ·
AI is not delegated. It is governed.
Last note in the series: the summary of the previous eleven in seven sentences, the decisions only management can take, the most used governance framework, the seven questions for the committee and a twelve-month agenda with a budget decided on numbers.
Read the note ›