Preface — Why This Book Exists
Why this book
Working with artificial intelligence has become strangely easy to do and strangely hard to do well. Anyone can open a chat window and get an answer in seconds. Getting answers you can rely on — accurate, consistent, worth using — is much harder.
We all tend to use the same handful of frontier models, so it is not the model that separates good quality from slop. The method matters — the practised habits by which an experienced person turns a capable tool towards a dependable result, the way a practised hand and a beginner, given the same tools, turn out very different work. This book is about that method: a structured way of working with AI, where you start from what you actually want, improve the result over several rounds, and check it before you use it.
A quick word on the terms, since they move quickly. When I say artificial intelligence, I mean the present generation of large language models — systems trained on enormous quantities of text, and increasingly images and sound, that respond to plain-language requests with fluent prose, working code, and structured analysis (Zhao et al., A survey of large language models, 2023). The most capable are called frontier models: the handful of largest, most general systems from a few well-resourced labs, the ones that set the pace and that everyone else measures against (Anderljung et al., Frontier AI regulation: Managing emerging risks to public safety, 2023). ChatGPT, Claude, and Gemini are the familiar names; behind them sits a foundation model, a single large network trained once at great expense and then adapted to countless tasks (Bommasani et al., On the opportunities and risks of foundation models, 2021).
A few terms used throughout, defined plainly:
- Large language model (LLM) — a network trained to predict the next word, which in scale yields fluent prose, code, and analysis.
- Foundation model — one large model trained once, then adapted to many tasks.
- Frontier model — the largest, most general foundation models that set the pace (ChatGPT, Claude, Gemini).
Why I wrote it
A few words about me. My name is Chris Tham. I run a boutique strategy consulting company, Hello Tham, and I also teach technology and information systems to undergraduate and master’s students at Torrens University Australia.
This book grew out of my recent experiences using AI in my work, and documents what I have learnt. I had just finished delivering an AI strategy for a global client, and I was also teaching AI to my students. Consulting gives me the experience, and teaching makes me explain it. This book is the written version of that explanation.
Why now, and won’t it date?
Why write this in 2026? Because method now matters more than the tools. Most organisations have AI somewhere in the building, yet few can point to real results rather than just activity (Stanford HAI, The AI index 2026 annual report, 2026; McKinsey & Company, The state of AI, 2025). And as the models improved, which one you use stopped being what sets results apart. What separates good results now is the work you do around the model: how you set up the task, what context you give it, and how you check what comes back.
That is also why the book should outlast its examples. The field moves in weeks, and a book of prompts and tool tips would be stale a model release later. This one rests on a steadier layer — how the models work, not which model leads this quarter — and those properties have held across generations of models. Where a point is tied to a 2026 product or figure, I mark it as such; the way of working around it is meant to last.
Who it is for
This book is for the thoughtful professional — a leader, a consultant, an analyst, a designer, a builder — who wants structured, effective use of AI rather than a bag of prompts. I assume
- you are numerate and can read a diagram, a code snippet, or a research paper when it helps;
- you have already used these tools, and seen both what they can do and how often they get things wrong;
- and you want to use AI well, not to build models.
None of this is essential, but a little more helps: some feel for how software and information systems are built and used, and enough Python or JavaScript to be able to review code generated by AI. You don’t have to work in IT to get a lot from it.
Using AI well is a discipline, and it is the one this book teaches. My promise is practical: an approach you can start using tomorrow, and that keeps improving with practice.
How to read it
The book is six chapters, and they build. The first lays the foundations — what AI is, how it works, and what it cannot do. The middle chapters climb from personal productivity, to building software, to the disciplines that keep that work sound, then to responsibility and governance. The last turns to mastery — what stays human when the tools are this good.
| Chapter | Theme | The question it answers |
|---|---|---|
| 1 | Foundations | What is AI, what is it not, where do we stand? |
| 2 | Productivity | How does it change individual knowledge work? |
| 3 | Software | How does it change building software? |
| 4 | Disciplines | What keeps that work sound at scale? |
| 5 | Responsibility | How do we govern it safely and fairly? |
| 6 | Mastery | What stays distinctly human? |
Read it once in order, to see how the ideas rest on one another, then come back for the parts you need. Every claim is cited inline to a primary source, so you can follow the trail yourself; where the evidence is thin, I say so.
How this book was made
This book is built the way it argues you should work, so it is worth a short note on how. Every chapter is a single plain-text file written in GitHub-Flavoured Markdown — the same lightweight format the models read and write best — with the diagrams drawn in Mermaid, a text notation for flowcharts and timelines. There is one source. The website you may be reading, the PDF, and the ePub are all generated from those files; nothing is laid out twice by hand.
The look comes from a single file of design tokens. Every colour, type size, and per-chapter accent lives in one theme.yml, and a small script turns it into the separate stylesheets each format needs — so adjusting a heading weight or a shade of pink is one edit that flows to the website, the PDF, and the ePub at once. The palette is Rosely, a set of soft, Pantone-named tones: a near-black ink it calls black-beauty, radiant-orchid for accents, raspberry-sorbet for links, chosen to stay calm across a long read. The type pairs Raleway for headings — its heaviest Black weight for titles — with Noto Serif for the body and Noto’s Japanese cut for the two kanji, 愛 and 道, that give the book its name. Both families are open-licensed and travel with the book, so it renders the same on any machine.
The build is Quarto. It renders the Markdown to a searchable website and, through the Typst typesetting engine, to a six-by-nine-inch PDF laid out like a printed trade book, and to a reflowable ePub for e-readers. The Mermaid diagrams are tinted to the same Rosely palette in every format, so a flowchart looks the same on the web, on the page, and on an e-reader. Push a change to the repository and a GitHub Action runs the same pipeline and publishes the result to the web.
I designed the cover myself, using AI to help create it in SVG — a text format a model can write directly. The design has three parts: the title AI-dō set in Raleway Black, a crumbling, glitching “AI” in the display face Rubik Glitch Pop, and, sweeping over both, the character 道 in KokuryuSou, a brush font of raw, dragon-like strokes. A heart-shaped 愛 is hidden inside the “A”. According to Chinese astrology, I am a wood dragon. The lettering is then flattened to vector outlines, so the finished file needs none of those fonts to display; a headless browser turns it into the cover image, and the same artwork, re-laid out wide, becomes the card you see when the book is shared online.
The Japanese motif is not confined to the cover; it runs through the whole book, and it is personal — the fruit of three years spent learning the language and immersing myself in the culture. The title is the seed: AI-dō, where 道 (dō) is the way, the lifelong disciplined practice behind arts like judō, and 愛 (ai) is love, the care for the people the work touches. Both characters, brushed in that same dragon-stroke hand, open every chapter over a seigaiha wave pattern drawn from classical Japanese design. The theme reaches into the argument as well. Chapter 2 is subtitled 愛 in practice, and Chapter 6 closes on shuhari (守破離) — the martial-arts progression from keeping the form, to breaking it, to leaving it behind.
Much of this toolchain was built the way the book describes: I said what I wanted, let the tools work out how, and checked what came back. The whole toolchain is open source, alongside the text, at github.com/ChristineTham/aidou.
References
Anderljung, M., et al. (2023). Frontier AI regulation: Managing emerging risks to public safety. arXiv. https://arxiv.org/abs/2307.03718
Bommasani, R., et al. (2021). On the opportunities and risks of foundation models. arXiv. https://arxiv.org/abs/2108.07258
McKinsey & Company. (2025). The state of AI. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
Stanford Institute for Human-Centered AI. (2026). The AI index 2026 annual report. Stanford University. https://hai.stanford.edu/ai-index/2026-ai-index-report
Zhao, W. X., Zhou, K., Li, J., Tang, T., Wang, X., Hou, Y., Min, Y., Zhang, B., Zhang, J., Dong, Z., et al. (2023). A survey of large language models. arXiv. https://arxiv.org/abs/2303.18223