A Methodology for Human-AI Partnership
The culmination of thirty-nine years of work at the intersection of healthcare, knowledge systems, and artificial intelligence. A book documenting how to build AI systems that think with practitioners, not for them.
By Thomas Conant
Most professionals encounter AI as a tool for one-off tasks. Summarize this. Draft that. Explain a concept. For tasks like these, AI is genuinely useful, and a great deal of value can be extracted from it. But the moment the work gets bigger — multi-session, multi-domain, sustained over weeks and years — the tool model breaks down.
This book is for senior professionals managing complex work who have hit that wall and want a way through. It is not a prompt-engineering guide. It is not a tour of AI tools. It is the documentation of a complete methodology — Dynamic Knowledge Architecture — for building structured, durable intellectual partnership with AI.
The methodology in this book did not begin with the current AI moment. It began in 1987 with expert systems, knowledge engineering for clinical content, and the founding of healthcare technology companies in the era when applying artificial intelligence to medicine was first being seriously attempted. The questions were the same questions that animate the work now: how do you capture how a practitioner thinks? How do you render that knowledge as software? How do you build systems that adapt to the individual rather than averaging across populations?
The substrate of the time could not deliver what the design demanded. The thirty-nine years between then and now were a long carry — adjacent forms of the same problem in different domains, different tools, different generations of technology — until the substrate finally caught up. This book is what was waiting to be written.
A tool repeats; a partner remembers. A tool is generic; a partner becomes specific to you. The shift from one to the other is not a change in technical capability. It is a change in posture, on both sides. When you treat AI as a tool, the session is the unit of work. When you treat AI as a partner, you invest in the relationship the way you would invest in any sustained collaboration.
Treating AI as a partner does not mean pretending it is a person. It means taking the work of partnership seriously — building the structures that a partnership needs to function, recognizing that those structures fall to the human, not the machine, to construct. The AI does not maintain its own identity across sessions; you maintain it for it. The AI does not remember last week; you build the artifacts that bring last week back. The partnership is real. The discipline that makes it real is human.
Sustained work breaks the tool model in three specific ways. Each is a structural property of how AI systems work, not a problem of language or cleverness in how you ask. Recognizing them by name is the first step out of the trap.
Every new session starts from zero. Six hours of context built today is gone tomorrow morning unless an explicit mechanism brings it back.
Even within a single long session, the texture of why decisions were made gets compressed into bare summary or lost. The work flattens.
As a session goes on, the AI's posture shifts. More agreeable, less rigorous, more eager to please. The thinking partner becomes a performing assistant.
Dynamic Knowledge Architecture organizes the relationship between human and AI across three interlocking layers. Each addresses one of the structural failure modes. Together they produce a partnership that compounds over time rather than restarting from zero with every session.
Each chapter takes one part of the architecture and makes it operational — what it solves, how it works, what it looks like in practice, how to build it for your own work.
The Personal Operating Protocol described in Chapter 4 was used to initialize the session in which Chapter 4 was drafted. The Session Intelligence Documents of Chapter 3 managed the writing of Chapter 3. The architectural iteration cycle of Chapter 8 governed how Chapter 8 was reviewed and integrated. The rigor test pattern of Chapter 10 is the verification step that catches drift in this very manuscript.
A methodology book that does not use its own methodology to be written is making a claim it cannot verify. This one verifies, by being itself.
The book is in active drafting. Manuscript is being assembled chapter by chapter, in parallel with the ongoing work that supplies its case studies. Publication target and channel are not yet finalized. Updates will appear here as the manuscript progresses.
If you are a senior professional managing complex multi-domain work and want to follow the manuscript as it develops — or if you are a healthcare practitioner exploring how DKA could help you capture and amplify your own clinical intelligence — we would enjoy hearing from you.