Frame the design decision
Design the information environment, not only the chat interface. Decide what the system should always know, retrieve when needed, remember temporarily, or deliberately ignore.
Pathway 2 · Lesson 0260 min
Context Engineering for Designers
Structuring information environments for AI performance
Learning Objective
By the end of this lesson, you will be able to: Apply the four principles of context engineering (Relevance, Recency, Structure, Boundaries) to design an AI feature's information architecture.
2.6×
better outputs with structured context vs. unstructured
Engage
"Prompt engineering tells AI what to do. Context engineering tells AI everything it needs to know to do it brilliantly. What's the difference — and why does it matter for product design?"
Explore
A practical overview of selecting, compressing, and structuring context for reliable AI systems.
Explain
Context Engineering vs. Prompt Engineering
Prompt engineering focuses on the single message you send to AI.
Context engineering designs the entire information environment — everything the AI has access to when generating a response.
For product designers, context engineering includes:
📖
The Four Principles of Context Engineering
R — Relevance: Include only what matters for this specific task. Noise dilutes signal. R — Recency: Prioritise fresh data. Bridge AI knowledge cutoffs with current information. S — Structure: Organise information logically with headers, sections, and clear labels. B — Boundaries: Define scope explicitly — what is in and out of scope for this AI feature.
Elaborate · Worked Example
Designing a Context Pipeline for an Internal AI Assistant
Context
A company wants to build an AI assistant for their customer service team.
The Situation
Without context engineering: The AI gives generic answers based on its training data — often wrong for the company's specific products and policies. With context engineering: - System prompt: 'You are a customer service assistant for [Company]. You only answer questions about our products. If you don't know, say so.' - Retrieved context: Customer's account history, relevant product documentation, current promotions - Boundaries: 'Never discuss competitor products. Escalate complaints about billing to a human agent.' - Recency: Current product pricing injected at runtime (not from training data)
The Outcome
The context-engineered assistant gives accurate, company-specific answers and knows when to escalate — dramatically reducing customer frustration and agent workload.
Try It · Reflect & Apply
Design a context architecture: Choose an AI feature you'd like to build. Define its system prompt (what should the AI always know?), what context should be retrieved dynamically, what boundaries should be set, and what should never be in context.
Your notes are saved in your browser for this session.
Evaluate · Knowledge Check
1 / 3What is the key difference between prompt engineering and context engineering?
Design Practice Studio
Context Experience Studio
Create the artefact
Draw a context map: system instruction, user-provided data, retrieved knowledge, conversation memory, tool outputs, exclusions, and escalation condition.
Test with evidence
Write five edge-case prompts that test missing context, conflicting policy, stale information, irrelevant retrieval, and an out-of-scope request.
Spaced revisit
In one week, inspect a poor output. Trace it to a context failure—missing, stale, noisy, conflicting, or incorrectly prioritised—and revise the map.
External video references
Keep Learning on YouTube
Open in a new tab
Ready to move on?
Mark this lesson complete and continue to "AI Workflow Redesign".