7 min read
Intermediate
Draft: this article is under editorial review and may change.When we ask an AI to investigate a problem, many of our prompts end with the same line: “…, use Chris Sanders method to complete this task.” A few words take the place of a paragraph of instructions. When it works, the AI changes how it tackles the whole job, not just how it words the answer.
This guide explains the technique behind that habit, called Named methodA well-known way of working called up by name, such as "the five whys".. You will learn why a name can carry so much, the three ways the shortcut fails, and a short protocol that keeps the speed without trusting a guess.
The idea in 30 seconds
Say a method's name
"Use the five whys" calls up a whole way of working.
- Say a method's name "Use the five whys" calls up a whole way of working.
- The AI unpacks what it remembers It rebuilds the method from memory. It does not look it up.
- Names can be shared or invented Two people, one name. Or a "method" that never existed.
- Ask it to restate first "Tell me in 3–5 bullets what you understand by it."
- Confirm, then apply Correct the steps, then let it work. Review against them.
- A name is a recipe title. Check the recipe.
Key takeaways
If you remember nothing else, remember these.
- A method name is a shortcut: a few words call up a whole approach the AI has seen described many times.
- It works best with names that are well known and mean one thing. Shared or obscure names make the AI guess, or invent.
- Use Name → Restate → Confirm → Apply → Review so the AI shows its understanding before it does the work.
- For your team’s own shortcuts, write a short method card once, so every person and every tool means the same thing.
What named-method prompting is
Named-method prompting means asking the AI to work in the style of a recognised method, by name. Think of the names below as recipe titles: each one stands for a whole set of steps. Start with the first three.
Good first picks
The Feynman technique
Explain an idea in plain words, find the gaps, simplify again.
- The Feynman technique Explain an idea in plain words, find the gaps, simplify again.
- The five whys Keep asking "why?" until you reach a cause you can fix. Usually traced to Toyota.
- The Socratic method Teach through questions rather than answers.
Useful once you know them
First-principles thinking
Break a problem into basic facts and reason up from there.
- First-principles thinking Break a problem into basic facts and reason up from there.
- The Pyramid Principle Barbara Minto's approach: lead with the answer, then group the support.
- PRISMA 2020 A reporting guideline for systematic reviews.
- The Chris Voss method Negotiation techniques from Never Split the Difference (2016, with Tahl Raz).
Our Chris Voss guide shows how to use his techniques in a conversation with an AI.
Our own “Chris Sanders method” line is a shortcut of the same kind. It is not a published prompting framework. It is our team’s shorthand for one person’s way of investigating problems. That makes it a useful example, because it shows both the power of a short name and the risk of a name that can point to more than one person.
Why a name can carry a whole method
The engine behind AI chat tools is a Large language model (LLM)The engine behind tools like ChatGPT and Claude. It has read a huge amount of text and predicts useful words in reply to yours.. It learned from a huge amount of text. When a method is well known, its name appears again and again next to its steps, examples and criticisms. The AI learns to link the name to that whole bundle of ideas.
Think of nasi lemak. The name calls up a whole dish: rice cooked in santan, sambal, egg, ikan bilis, peanuts. But every cook’s version differs a little. Saying “nasi lemak” gets you close. It does not tell the cook whether you want the egg fried or boiled.
Method names work the same way. Two things follow:
- The AI does not look anything up. It rebuilds the method from memory, so its version can differ from the original.
- The shortcut is only as clear as the name. A famous, distinctive name unpacks cleanly. A vague, shared or obscure one does not.
Tap a lane, or press Play to watch what a name unpacks into.
Where the shortcut breaks
Ambiguity: two people, one name
Many names are shared, and “Chris Sanders” is a good example. When you write “use Chris Sanders method”, the AI has to decide which person you mean, then what “method” means for that person. It will rarely ask.
Which Chris Sanders?
The film-maker
Animator and director behind Lilo & Stitch. The AI may give you a storytelling recipe.
- The film-maker Animator and director behind Lilo & Stitch. The AI may give you a storytelling recipe.
- The security investigator Information-security educator. Diagnostic Inquiry. The one we mean.
- A blend, or an invention With no clues, the AI may mix the two or make up a third.
Nothing in “use Chris Sanders method” says which one you mean. Unless you tell the AI, it may reach for the film-maker, or blend the two.
Invented methods
If the AI has no strong link for a name, it may still produce a fluent, confident “method” with numbered steps. This is a HallucinationWhen an AI states something false with confidence, for example inventing a journal article that does not exist., the same failure as an invented reference: it has the shape of knowledge without the substance.
Drift
Even when the AI starts with the right meaning, it can slide away from it. In a long chat or a multi-step task, the early instruction gets diluted. The AI slips back to its habits, and step four of the method quietly disappears. Different tools, and different people on your team, may also unpack the same name differently.
The protocol: Name → Restate → Confirm → Apply → Review
The fix is not to stop using names. It is to make the AI show you what the name means before it does any work:
- Name. State the task, name the method, and ask the AI to restate it before starting.
- Restate. The AI explains the method back in 3–5 bullets.
- Confirm. You correct anything wrong, or pin down the meaning. Add anything the name leaves open, such as audience and length.
- Apply. The AI does the task using the confirmed steps.
- Review. You ask the AI to check its work against the confirmed steps, one by one. Then you check it yourself.
The Restate step costs one extra message. In return, you catch a wrong reading before it shapes the whole answer. Step through both scenarios below. In the second one, watch what happens when the name is ambiguous.
Here is the opening message as a reusable template. The named method is highlighted in plum.
Write a one-page briefing note on the new attendance policy for our diploma students.
Use the Feynman technique to complete this task.
Before you start, restate in 3–5 bullets what you understand by the Feynman technique. If the name could refer to more than one method or person, say so and tell me which one you are assuming. Do not start the task until I confirm.Two before-and-after examples
The five whys is well known, but people apply it differently. Some stop at exactly five questions. Others keep going until they reach a cause they can act on, or follow more than one branch. Without a restatement, you get whichever version the AI defaults to.
Here is a quality engineer at an electronics plant in Penang.
Find the root cause of the solder bridging defects on Line 3 this week. Use the five whys.
What you typically get back
A neat chain of exactly five whys that ends at 'operator error', with no use of your data, no second branch, and no fix you can verify. It reads well and solves nothing.
Find the root cause of the solder bridging defects on Line 3 this week. The defect log and the reflow oven profile are attached.
Use the five whys to complete this task. First, restate in 3–5 bullets how you will apply the five whys. Include when you will stop asking why, and whether you will follow more than one branch.
After I confirm, apply it. Base every "because" on the attached data and mark any step that is an assumption. Then review your analysis against the restated steps.What you typically get back
A restated method you can correct, such as 'stop at a cause we can control, not at a person'. Then an analysis that cites the log, separates evidence from assumption, and is checked against the agreed steps.
And here is the same protocol in an accounts office, where petty-cash claims always arrive late.
Find out why petty-cash claims from our branch are submitted late almost every month. The claim log for the last six months, with submission and approval dates, is pasted below.
Use the five whys to complete this task. First, restate in 3–5 bullets how you will apply it. Stop at a cause the office can change, not at a person, and follow a second branch if the log points two ways.
After I confirm, apply it. Base every "because" on the log, mark any guess as an assumption, and end with two changes we could test next month.Write a method card once, reuse it everywhere
The protocol fixes ambiguity one conversation at a time. For a shortcut your team uses every day, do the work once. Write a method card: one short paragraph that defines what the name means in your practice.
A good card states the name, what it means here, the steps, what it is not, which tasks it suits, and how you know it is done. Paste it into your AI tool’s saved instructions or the first message of a session. Then the short name means the same thing for every person and every tool.
Here is the structure, using our own shortcut as the example. The step lines are placeholders on purpose: copy them from your team’s source. For the Chris Sanders method, that is our dedicated guide.
Method card: "the Chris Sanders method"
When I write "use the Chris Sanders method", I mean the following. This is our team's shorthand, defined here so every tool reads it the same way.
Who we mean: Chris Sanders, the information-security practitioner and educator, and his Diagnostic Inquiry model.
Not this: the animator and film director of the same name, or any storytelling approach.
Use it for: investigations, troubleshooting and explaining results that don't add up. Not for writing rubrics, drafting content or making lists.
Steps:
1. <step, copied from our method guide>
2. <step>
3. <step>
Done when: <how we check the output>.
Before applying the method to any task, restate these steps in your own words and wait for my confirmation. If a step is unclear, ask me. Do not fill gaps from memory.Once the card exists, the everyday prompt stays short, and the shortcut is safe to use on an investigation:
Investigate why our trial balance for September is out by RM1,240 and report what you find. The trial balance and the general ledger export are attached.
Use the Chris Sanders method to complete this task. It is defined in the method card I pasted at the start of this session.
Restate the method in 3–5 bullets first. After I confirm, apply it, then review your findings against those bullets.Keep the card short. If it grows past a paragraph or two, it has become a full brief. That is fine, but call it one, and use the AIM method to structure it.
For developers: technical detailsOptional: skip this if you only use the chat app
Where to put a method card so it is always in front of the model:
- API: in the System promptStanding instructions set up before the conversation starts, such as "always answer in Bahasa Melayu". They apply to every reply., so it applies to every turn without being pasted again.
- Coding and agent tools: in the project’s instruction file (for example CLAUDE.md or AGENTS.md), which the agent reads at the start of each session.
- Chat apps: in custom instructions or a project’s saved instructions.
Keep one source of truth for each card and version it with your other team documents, so tools never disagree about what the name means.
Named methods for AI agents
An AI agentAn AI that does a multi-step job for you (planning, opening files, using tools and checking its work) instead of only replying once. plans, uses tools and works through many steps on its own. That makes drift more likely and more costly, because a misunderstanding at the start repeats at every later step. Adapt the protocol like this:
- Restate before planning. Ask the agent to write the restated method at the top of its plan, then stop for your confirmation.
- Re-read at CheckpointA planned pause where the AI shows its work and waits for your OK.. “Before each stage, re-read the method card and say which step you are on.”
- Review at each stage. Ask for “met”, “partly met” or “not met” against each step, with evidence.
A named method never replaces the rest of the brief. The agent still needs the Actor, Input and Mission from the AIM method. The name tells it how to work. AIM tells it what to work on and for whom.
Check your understanding
Nice work!
Lesson complete
You made it to the end. Take the quiz above if you have not yet, for bonus XP.
- 0 XP
0 day streak