The Workbook: Start Here

What this is

The Handbook explains the architect’s job. This is where you do it.

Ten exercises, one per Handbook module, each producing a real artifact for Lintel — a fictional but fixed company with a legacy monolith, no spare engineering capacity, and a failed ML project nobody mentions. By the last one you hold a complete architecture portfolio for a realistic engagement.

About 33 hours of work — nearly half the guide, alongside about 35 hours of Handbook reading.

Do these in order — they are not independent exercises

Each exercise consumes what the previous ones produced. Exercise 3 scores its design against the requirements budget from Exercise 2. Exercise 7 red-teams the designs from 4, 5 and 6. Exercise 9 reconstructs the reasoning behind decisions you made in 3 through 7. Exercise 10 assembles all of it.

Pick one at random and its inputs will not exist. If you are going to do some but not all, start at 1 and stop where you stop — a partial chain works; a scattered selection does not.

Optional, and the half where the learning happens

Reading the Handbook gives you recognition — you will spot the patterns and know the vocabulary. Working the exercises gives you capability, because they force the decisions the reading only describes.

The guide makes this argument about itself in the final module: a design you have reasoned about is defensible, and only a design that met reality is tested. These exercises are the closest a self-study guide can get to the second thing. They are genuinely optional, and skipping them changes what you get out of this.

Before you start

Read the corresponding Handbook module first. Each exercise assumes its module, and most reach back to earlier ones.

Read the Lintel case. Every exercise uses those fixed facts, and you should not invent new ones — where you need a detail the brief does not give, state it as an assumption.

Or substitute your own system, if it has all five of: a genuine business driver with a number attached, more than one stakeholder with conflicting interests, a legacy component you cannot replace, a compliance constraint, and a plausible AI component. Without all five, several exercises lose their teeth.

The ten exercises

1. The Architect's WorkA business-value brief, a system context diagram, a first stakeholder map and a charter. ~2 h.

2. From Business Need to RequirementA requirements catalog, an ASR list, a Utility Tree and an NFR budget. ~3 h.

3. Choosing a StructureA style decision, a reference architecture sketch and a sequenced roadmap. ~3 h.

4. The AI StackA stack selection, a RAG-or-fine-tuning decision, a data strategy and a serving topology. ~2.5 h.

5. Delegating to AgentsAn agent design, an action boundary, a tool integration, a handoff protocol and an observability plan. ~2.5 h.

6. Running It in ProductionA unit-economics model, a threat model, a privacy design and an evaluation plan. ~3 h.

7. Designing Under UncertaintyAn optimisation plan, an uncertainty policy, an antipattern review, and a red-team that yields your defect list. ~3 h.

8. Working Through the OrganisationAn organisational analysis, an influence map, a positioning statement, a justification pack, three framings and a pitch. ~5.5 h.

9. Making the Practice RepeatableAn ADR set, an HLD, a proposal template, a documentation structure and a process proposal. ~4.5 h.

10. Evolving What ExistsThe capstone: debt, legacy and migration plans, a gap analysis, a learning roadmap and the final portfolio. ~4 h.

How the chain runs

graph TD
  E1["<b>1 · The Architect's Work</b>"] -->|"business-value brief"| E2["<b>2 · Requirements</b>"]
  E2 -->|"ASRs, NFR budget"| E3["<b>3 · Structure</b>"]
  E3 -->|"reference architecture"| E4["<b>4 · AI Stack</b>"]
  E4 -->|"stack, serving topology"| E5["<b>5 · Agents</b>"]
  E5 -->|"agent design, action boundary"| E6["<b>6 · Production</b>"]
  E6 -->|"the Exercises 4–6 design"| E7["<b>7 · Uncertainty</b>"]
  E7 -->|"findings"| E8["<b>8 · Organisation</b>"]
  E1 -.->|"stakeholder map v0"| E8
  E8 -->|"influence map, justification pack"| E9["<b>9 · Practice</b>"]
  E7 -.->|"defect list"| E9
  E9 -->|"ADR set, HLD"| E10(["<b>10 · Capstone</b><br/>everything above"])

  classDef early fill:#5B8FB9,stroke:#4A7A9E,stroke-width:2px,color:#ffffff
  classDef mid fill:#E8A33D,stroke:#C4832A,stroke-width:2px,color:#ffffff
  classDef late fill:#7D9A4E,stroke:#5F7A38,stroke-width:2px,color:#ffffff
  classDef cap fill:#D9695B,stroke:#B4483A,stroke-width:2px,color:#ffffff
  class E1,E2,E3 early
  class E4,E5,E6,E7 mid
  class E8,E9 late
  class E10 cap

Every arrow is an artifact one exercise hands to a later one. The dashed arrows are the long reaches — the first stakeholder map is corrected seven exercises later, and the defect list skips one before becoming decision records. The capstone consumes all of it.

How to work one

1. Read the Handbook module. The exercise assumes it.

2. Check the inputs. Each page opens by naming which earlier artifacts it needs.

3. Timebox it. The estimates above are deliberate — these are judgement exercises, not polish exercises.

4. Score yourself against the rubric. Every exercise states one, plus a self-check list. The rubric is the point: it tells you what “done well” means before you start.

→ Begin with The Architect’s Work, after reading its Handbook module.