Preface
In this book
Beyond the primary audience in Biometrics, colleagues in adjacent roles, including clinical operations, data management, quality, and regulatory affairs, will find the same principles useful wherever they own or review a comparable reporting process.
After reading the Introduction, you will be able to:
- distinguish AI adoption for an isolated task from an AI-first workflow;
- state the five principles of trustworthy assistance and read a task contract;
- name the six stages of the AI-first software development lifecycle (AI-first SDLC), their canonical artifacts, and the human gate at each stage; and
- choose one bounded workflow in your own work that could become AI-first.
Familiarity with clinical study artifacts such as the protocol, statistical analysis plan, and tables, listings, and figures will help you connect the examples to your own work. The main route requires no coding experience and no prior experience building AI agents. The optional technical route uses R.
How the book is organized
The book has two parts:
- Introduction defines the shared concepts once: what an AI-first workflow is, what an AI agent is, the principles of trustworthy assistance, the task contract, and the AI-first SDLC that the examples share.
- Applied examples develops bounded workflows through that lifecycle. Each example states its teaching purpose, maturity, and prerequisites up front, and the examples are ordered by increasing maturity. The rounding example (6 Rounding in analysis and reporting and 7 From one-file skill to maintainable rounding workflow) is the most complete; later examples follow the same lifecycle at the maturity their evidence supports.
How to use this book
The main route is reading: begin with the Introduction, then follow the applied examples in order.
- Main route (no setup required). Read the chapters in order and work the exercises through a browser-based AI agent. You will specify requirements and review prepared code and results. The exercises supply what the agent needs, such as the synthetic dataset in the day-one exercise used by 3 Onboarding an AI agent, so no local R, Python, Git, or Codespaces setup is required on this route.
- Technical route (optional). To work in a repository rather than a browser, open the book repository locally or in Codespaces. The hands-on exercise is 8 Lab: Run a one-file rounding skill, which reviews a synthetic R example against one business rule and compares the response with a published answer key. That chapter supplies the instruction file the exercise uses and states what the result does and does not establish. An available R installation lets the agent run the small numeric probe the lab describes. Contributor setup is documented in the repository README.
- Facilitator route. Participant worksheets, worked facilitator examples, and the assessment rubric are in
workshop/and are linked from the example chapters that use them. Reusable artifact templates are intemplates/.