LUDA Research Group

Why a 'Good Environment' Matters More Than a 'Good Question' for AI

Have you ever used ChatGPT or Claude and walked away disappointed with the results? The quality of an AI's output depends just as much on what kind of environment you give it to work in as it does on how you phrase your question. This is the core idea behind Context Engineering — the concept that has taken the global AI community by storm since the second half of 2025.

In June 2025, Shopify CEO Tobi Lütke first defined the concept: 'The art of providing all the context a large language model (LLM) needs to plausibly complete the task at hand.' Just one month later in July, Gartner — the world's largest IT research firm — made a bold declaration: 'The era of prompt engineering is over. Context engineering is what comes next.'

'Asking Better Questions' vs. 'Designing a Better Environment' — What's the Difference?

Prompt engineering is the craft of figuring out how to ask an AI the right questions. Think of it as a collection of tips and tricks for phrasing things in ways that get better answers. Context engineering, on the other hand, is a much broader concept. It's about designing the entire environment in which your AI operates.

Context refers to the full package of information delivered to an AI:
- Role instructions (system prompt): "You are a customer service specialist."
- Conversation history: "This customer previously requested a refund."
- Reference documents: "Here is our company's refund policy."
- Available tools: "Calendar booking, email sending."
- Response format: "Keep it under 3 sentences, in a friendly tone."

Context engineering means systematically assembling all of these elements and providing them to the AI. Here's a simple analogy: if prompt engineering is giving a new employee clear task instructions, then context engineering is setting up their desk, handing them the right tools, providing reference materials, and creating the ideal work environment so they can do their best work.

Research backs this up. According to an AI agent comparison study from 2026, agents operating without context files had a task success rate of just 30% — while agents given well-designed context achieved success rates of up to 90%. Same AI, same tasks — but a 3x improvement just from better environment design.

Where Are New Zealand and South Korea in This Journey?

[New Zealand] AI adoption is spreading fast, but knowing how to use it well is still a work in progress. As of 2026, 87% of New Zealand businesses are using AI in some form (up sharply from 48% in 2023), and 91% of workers report using generative AI. Among businesses using AI, 91% say it has improved operational efficiency, and 77% report reduced operating costs.

However, only 24% of New Zealand adults have received any AI-related training, and 43% of those who haven't started using AI cite a lack of expertise as the main barrier. This is precisely where context engineering becomes so valuable — it's the most practical first step for moving beyond just using AI to actually using it well.

[South Korea] As of 2025, 55.7% of South Korean businesses are using generative AI, with projections suggesting that figure will climb to 85% by 2026. Major conglomerates like Samsung, LG, and Hyundai Motor are already applying context engineering to AI coding tools to boost developer productivity. SK Telecom's developer community platform, DEVOCEAN, has also highlighted context engineering as a core AI competency. At the International Conference on Machine Learning (ICML 2026) held in Seoul, approximately 25% of workshops focused on agentic AI — and context engineering was front and center as a key topic.

Context Engineering You Can Try Right Now

Don't overthink it. Compare these two approaches:

Standard prompt:

'Write me a marketing email.'

With context engineering applied:

'Our company runs an online education platform targeting working professionals in their 30s and 40s in New Zealand. We just launched a new introductory AI course this month. Write an email in a friendly, casual tone to existing subscribers who've received our newsletter before — encouraging them to sign up. Keep it under 200 characters.'

The second approach produces far more specific and useful results. Providing the role, audience, purpose, and format upfront — that is the most fundamental principle of context engineering.

Key Takeaways

Context engineering goes beyond crafting good questions — it's the discipline of designing the entire information environment in which your AI works.
The concept was defined by Shopify's CEO in June 2025, and Gartner declared it the core technology replacing prompt engineering.
Research shows that well-designed context can boost AI task success rates from 30% to 90% — a 3x improvement.
New Zealand has high AI adoption (87%), but training in how to use AI effectively still lags behind.
In South Korea, major companies including Samsung, LG, Hyundai Motor, and SK Telecom are already integrating context engineering into real-world workflows.

Wrapping Up