Before you begin — about this guide

— Question, collect, analyse, communicate: speeding up the loop that moves research and practice —

What this guide is for

Foundations covered meeting generative AI, the differences between the major tools, prompt patterns, and GPTs and Gems, and gave you a feel for AI as intellectual training wheels. Working with Your Own Files moved from chat to workspace: handing files to AI, growing a knowledge base, and reaching the doorway of AI operating files itself. Coding and Building Systems took you from user to maker, as far as building your own site and publishing it to the world.

This guide continues from there. Its theme is

learning to handle the data that moves research and practice — quickly and correctly — alongside AI

Concretely: the whole journey of data — framing a question, collecting, analysing, communicating — sped up with generative AI as your partner.

One thing needs saying plainly. This is not simply a book about running statistics with AI. Precisely because you can turn the crank quickly, the direction you are turning it in comes into question. As a compass, this guide brings in the footing that research has spent a long time polishing: epidemiology, social science and causal inference.

AI makes calculation, charting and drafting astonishingly fast. But get "what am I asking?", "does this number mean anything?", "am I mistaking correlation for causation?" wrong, and that speed becomes speed in a straight line towards a wrong conclusion.

AI makes analysis fast. Epidemiology and causal inference keep analysis from going wrong. Holding both is the goal of this guide.

And there is something to say to therapists at the outset. Data analysis is not a foreign field. Assessing a patient, forming a hypothesis, testing it, confirming the effect — what you do every day in the clinic is data thinking itself. This guide is about carrying those clinical muscles into research and practice.

What you should come away with

• A map for decision-making that moves from instinct to data, and from data to meaning
• A clear line on ethics and security in research and survey data, so you can move forward without unease
• The ability to frame a good question before analysis — induction, deduction and abduction, run with AI
• The ability to design what to measure and how, through the lens of epidemiology (study design and variable types)
• The ability to collect and organise both quantitative and qualitative data with AI
• The ability to design surveys with AI in a form that carries less bias
• The ability to work from univariate to multivariate analysis with AI as a partner — and to know its limits
• The ability to spot, yourself, the biggest pitfall of all: mistaking correlation for causation
• The ability to turn numbers into output that lands — figures, reports, slides, presentations, business proposals
• And above all, sight of the line between what may be handed to AI and what a human must carry

Who this is written for

• Therapists who have finished Coding and Building Systems and want to move data through research and practice
• People who hold clinical data, surveys and assessment measures but are not making use of them
• People who want to work on conference presentations, papers or case research (or already do)
• People going independent or running a community, who want to shift decisions from instinct to data
• People with an eye on joint research with universities or on building evidence
• People who have avoided analysis because "statistics is not for me"

If you already work professionally with statistical analysis and study design, much of this will feel basic. It is written as a bridge across to the side that works with data.

You can follow this guide without the earlier parts, but the thinking on security and ethics continues from Chapter 2 of the previous part and Chapter 3 of the one before. If you feel uneasy, reading those first will help.

How to read it

• Read from Part 1 in order. It is built to accumulate along the flow of question → collect → analyse → communicate.
• Each chapter ends with a "try this" and a summary. Read with your hands moving on data of your own — starting from anonymised or invented data, without exception.
• Statistical terms appear, and all of them are unpacked. There is no need to memorise. The goal is knowing that a way of thinking exists, and being able to pull it out with AI when you need it.
• The features, limits and prices described are as of the time of writing (2026). AI moves quickly, so details will change. The underlying way of thinking changes far more slowly.
• And one request above all others. Wherever responsibility is involved — research, conference presentations, papers, business decisions — always have AI's output verified by a specialist: a statistician, a supervisor, the original sources. This guide is a tool for making that verification fast and intelligent. It does not stand in for the judgement itself.

How it is organised

Five parts.

• Part 1 | Foundations — Why a therapist becomes someone who works with data. And the ethics and security that make it safe.
• Part 2 | Questions — Before analysis. Designing what to look at, through hypothesis thinking and study design.
• Part 3 | Collecting — Gathering data through two kinds of attention: quantity and quality.
• Part 4 | Analysis — Reading it with AI. And not mistaking correlation for causation.
• Part 5 | Communicating — Turning numbers into something that lands, and connecting to research, practice and publishing.

Let us begin.