TECH2030
Insights
AX Strategy · 3 min readAI-assisted

Harness Before Prompts: Write Down 'Your Standards' First

QJC Live covered the idea that writing down your standards, context, and tone—the 'harness'—comes before prompt skills.

Key Summary

  • Harness = rules documenting your standards, context, and tone
  • Self-objectification comes before prompt techniques
  • Separate task automation from generating new concepts
  • Can your team's standards fit on a single sheet of A4 paper?

The idea that having a good conversation with AI requires writing down your own standards, context, and tone before worrying about prompt techniques was the subject of a 128-minute livestream published by QJC Quantum Jump Club on its YouTube channel on October 2, 2026. The guest was Professor Kim Kyung-mook, who came from Samsung Electronics' design management division and has taught reasoning based on humanities and design thinking.

What Is a Harness, and Why Does It Come Before the Prompt?

The harness introduced by the channel is a set of working rules in which you write down your standards, context, and tone in advance so you don't have to explain them to AI every time. The channel explained that to write this, you first need to know who you are, which is why the order places it ahead of the prompt. The livestream discussed how, viewed from a humanities and human-centered perspective rather than a technical one, the harness represents the user's identity.

Professor Kim's class includes the following exercise.

  1. Students are each given two sheets of A4 paper.
  2. On one sheet, they draw their own face; on the other, they write a description of their own face in words.
  3. Each person then explains their work.

In a message sent on October 1, Professor Kim said, "I believe this kind of training is harness technique, and it's important that it happens before learning prompts."

How Should You Explain Things to AI?

The principle presented in the livestream is that when conversing with AI and building a harness, you should not try to persuade the AI, but instead objectify what you know and explain it concretely. In the video, the statement appeared: "Don't try to convince the AI. Just explain only as much as you actually know."

The same livestream discussed how the essence of design lies in establishing a clear concept in language and then realizing a form that fits that concept. It also noted that Stanford University's five-stage design thinking process doesn't function just by following the methodology alone—the environment and cultural foundation must be in place first for it to work. It was also mentioned that a distinction should be made between task automation, which handles repetitive manuals, and AI-based creativity, which generates new concepts.

Implications for AX at Korean Companies

If your company's in-house AI training starts with how to write prompts, it may be worth reviewing the sequence. If a harness consists of rules documenting an individual's standards, context, and tone, then at the organizational level, the proper order is to first check whether the team's criteria for judgment and work context can be put into writing.

When importing external AI adoption procedures, it is also safer to first examine whether the organization has the environment in which that procedure can actually function. By separating tasks with repetitive manuals as automation challenges from tasks requiring new concept generation as AI-based creativity challenges, you can set separate expectations for each type of task.

This week, test whether your team can write down, on a single sheet of A4 paper, the standards and context you repeatedly explain to AI every time.

Source: To Have a Good Conversation with AI, Try Writing Down Your Own Face Before the Prompt | Professor Kim Kyung-mook

If you found this helpful, share it.
Related servicesAX Transformation

Related insights

More in this category
Enterprise AX
Want to apply this to your own operations?
Request an AX consultation