19 - Solutions Architecture & Communication

The rest of the course teaches how to build GenAI systems. This module covers the work around the build that decides whether anyone funds, trusts or uses them: finding the real problem, qualifying it and putting a number on its value, writing down the architecture and its trade-offs, explaining it to engineers and executives in their own terms, and taking a pilot into production. Most failed AI projects fail here, not in the model.

Learning objectives 5-6 hours

By the end of this module you will be able to:

Run a customer discovery conversation that surfaces the workflow, the data, the success metric and the constraints before any solution is proposed
Qualify an AI use case on value, feasibility and risk, and build a cost and ROI model a finance reviewer would accept
Write an architecture document with C4-style diagrams, ADRs and an explicit trade-off table
Present the same system to a technical and an executive audience, including uncertainty and eval results, and handle the common objections
Plan a pilot with exit criteria agreed up front, and take it to production with adoption metrics and a handover

Prerequisites

Any build module, ideally RAG or Production Agents - you need a system to architect
Building Your Own Evals - success metrics are evals

Where This Module Fits

flowchart LR
    D["🔍 01 Discovery<br/>problem, workflow, data"] --> Q["⚖️ 02 Qualification<br/>& ROI"]
    Q --> A["🏗️ 03 Architecture<br/>docs & ADRs"]
    A --> C["🗣️ 04 Communicating<br/>to every audience"]
    C --> P["🚀 05 PoC to<br/>production"]
    P -->|"measured outcome"| D
    A --> SD["📐 System design:<br/>discovery to architecture"]

    style D fill:#d8dfe8,stroke:#b0bac8
    style Q fill:#e8e0d4,stroke:#c8b89a
    style A fill:#dde4dc,stroke:#b0c4b0
    style C fill:#ddd8e4,stroke:#b8b0c8
    style P fill:#e8e2d9,stroke:#ccc4b8

Chapter Map

#	Chapter	You will learn	Time
1	Customer Discovery for AI Projects	Why projects fail at the problem stage, discovery interviews, mapping the workflow, data and process readiness, success metrics	45 min
2	Use-Case Qualification & ROI	Value/feasibility/risk scoring, when not to use an LLM, unit cost model, ROI with sensitivity, build vs buy vs fine-tune	50 min
3	Architecture Docs & ADRs	Design-doc structure, C4 levels, non-functional requirements for AI, ADRs, trade-off tables, risk register	45 min
4	Communicating to Technical & Executive Audiences	Answer-first structure, one system told two ways, explaining uncertainty, demos, objection handling	40 min
5	PoC to Production	PoC vs pilot vs production, exit criteria, staged autonomy, adoption metrics, change management, handover	45 min
6	Q&A Review Bank	16 questions across the module	30 min

System Design

#	Case study	Domain
1	Discovery to Architecture	Accounts-payable invoice processing at a distributor - from discovery notes to architecture, ADRs, ROI and a pilot plan

Resources

Module quiz - every Check Yourself question in this module
Readiness Self-Assessment - this module covers the "customer discovery and architecture communication" domain
Capstones - every capstone now asks for an architecture document, ADRs and a business outcome

Key Cross-References

Cost models in depth → Cost and Latency, Prompt Caching & Cost
Choosing a cloud platform → Platform Comparison Table
Rollout and rollback mechanics → Model Lifecycle & Rollout
Operating what you ship → LLM Observability, SLOs & Incident Response

Section Appendix

Summary & Key Terms - a quick recap of this section and its essential vocabulary.

Next Topic

Previous: 18 - Agent Engineering · Next: Capstones