Stage 1 set out to build the conceptual foundation a practitioner needs before beginning applied work with AI tools. The argument the four modules developed together is that the current generation of AI tools works well enough for general knowledge work for the first time in the field's history, that the underlying mechanism explains both what the tools can do and how they fail, and that a practitioner who understands both can begin using these tools productively while retaining the professional judgment their work requires.
The through-line across the four modules is that AI operates as a specific kind of system with specific properties that shape specific professional practice, rather than as an abstract technology to be understood in general terms. The properties of a large language model, the ways it processes language, the boundaries of what it knows, and the patterns by which it fails carry practical weight rather than reflecting incidental technical detail. These properties determine what a practitioner can rely on the system to do, what they need to verify, and how they should structure their own work to take advantage of the system's strengths while catching its weaknesses. Understanding the properties is the foundation for everything else this programme will develop.
Module 1.1 established where the current generation of AI came from and why it works now. Three waves of AI research (symbolic systems in the 1950s through 1980s, machine learning from the 1990s through the 2010s, and transformer-based systems from 2017 onward) each produced real capability and hit different limits. The transformer architecture of the current wave is different because it scales, meaning that capability continues to improve as models get larger and training data grows. The AI tools a practitioner uses today are products of this scaling, and the reason they suddenly work for general knowledge work traces directly to the architectural breakthrough of 2017 and the years of scaling that followed.
Module 1.2 developed what AI actually is as a system. The four-layer framework distinguishes automation (fixed rules executed without learning), rule-based AI (explicit decision logic), machine learning (patterns learned from data), and deep learning (multi-layer neural networks, including the large language models that dominate current professional use). A large language model is a very large neural network trained to predict the next token in a sequence of text. Its capabilities, its quirks, and its failure modes all trace back to this underlying mechanism. Key terms the practitioner will encounter across their working life (token, context window, inference, hallucination, fine-tuning, retrieval-augmented generation) were defined at mechanism level in plain language, not as a glossary of jargon but as concepts a practitioner can actually work with.
Module 1.3 surveyed how AI shows up in professional work across the six main domains this programme covers (consulting, legal, insurance, finance, commercial real estate, and residential real estate) and the two auxiliary domains (marketing and business operations). Each domain has specific tasks where AI currently adds substantial value, specific mechanisms by which it does so, and specific limits that shape how it should be used. The three-level framework for human-AI collaboration (assisted work, augmented work, and autonomous systems) captured the patterns that recur across all eight domains. The surveys also revealed that AI is unevenly distributed across professional work, that the value of AI depends substantially on how it is integrated into practice rather than on the raw capability of the underlying technology, and that the failure modes of AI differ from the failure modes human professional work produces.
Module 1.4 converted the conceptual foundation into applied capability. A practitioner who has worked through this module can see their own work as a system of tasks rather than as a role description, construct well-formed prompts using the five-component structure (role, task, context, format, constraints), progress through the levels of prompting skill from simple prompting to structured prompting to more advanced patterns, apply working frameworks for common situations (CREST for general tasks, SCQA for analytical work, DEEP for refinement), iterate productively when first outputs need improvement, recognise the five failure patterns every practitioner should know (hallucination, confident imprecision, context limits, stale knowledge, pattern over-application), apply the verification discipline that catches those failures, and build the sustained practice through which AI capability develops over time.
What Stage 1 did not cover is the structured collaboration discipline that serious professional AI use requires when AI-generated work moves into decisions, client deliverables, and operational commitments. Stage 1 taught a practitioner how to work with AI. The integration of AI into professional practice at the level where accountability, defensibility, and governance become central remains Stage 2's task. The collaboration disciplines Stage 2 develops (the protocol of interrogation for evaluating AI output, the triangulation discipline for independent verification, the reasoning control that keeps outputs as inputs to human judgment, and the decision ownership that remains human responsibility regardless of how much AI-augmented analysis supports the decision) all build on the Stage 1 foundation. A practitioner who does not understand what AI is, how it shows up in their work, how to use it productively, and how it fails will find Stage 2's disciplines abstract and burdensome. A practitioner who has absorbed Stage 1 will find those same disciplines practical and necessary.
The shift from Stage 1 to Stage 2 is the shift from "understanding AI" to "collaborating with AI professionally." The technical capability to produce AI-augmented work differs from the professional capability to produce defensible AI-augmented work, and the distance between the two is where many AI deployments fail in practice. Stage 2 develops the professional capability. Stage 3 and Stage 4 then develop the applied practice within specific professional domains, drawing on both the conceptual foundation of Stage 1 and the collaboration disciplines of Stage 2. Stage 5 develops the governance and leadership dimensions that organisations adopting AI need to think about at the organisational and cross-functional level.
A practitioner reading this summary has now completed the foundational stage of this programme. The conceptual understanding is in place. The applied capability is in place at the level Stage 1 set out to produce. What remains is the progressive development of professional-grade practice, which the stages ahead will build one layer at a time. The work that follows is easier than the work that has been completed, because the hardest part of learning to work with AI professionally is establishing the mental model that makes everything else make sense. That mental model is now established. Everything that follows builds on it.