AI한테 일만 시키지 말고, 나를 점검시키세요Don't just put AI to work — put it to work on you.
제가 제 AI 운영을 하나씩 뜯어보며 골라낸, 나를 바꾸는 질문 10개.
그대로 복사해서 쓰세요.
10 questions I pulled from auditing my own AI setup — copy them and
use as-is.
한 · 영 병기복사해서 바로 사용ChatGPT · Claude · Gemini
대부분은 AI한테 “이거 해줘”라고 시킵니다. 결과물은 나오는데, 내 생각은
그대로예요. 진짜 차이는 “내가 뭘 놓치고 있는지 찾아줘”에서 납니다.
아래 10개가 바로 그런 질문이에요.
Most people tell AI “do this.” You get output, but your thinking
stays the same. The real difference comes from “find what I'm
missing.” These 10 are exactly that kind of question.
저는 제 사업과 AI 운영 전체를 이 질문들로 하나씩 뜯어봤고, 그중 누구의
일·커리어에도 그대로 통하는 것만 골라 10개로 추렸습니다. 쓰는 법은
간단해요. 복사해서 붙여넣고, 대괄호
[ ]
안만 당신 상황으로 채우면 됩니다.
I put my own business and AI operation through every one of these,
then kept only the ones that transfer to anyone's work or career —
ten of them. Using them is simple: paste, and fill the brackets with
your own situation.
왜 이게 다른가 Why these hit different
1
감사시켜라
Audit, don't command
질문의 방향을 결과물이 아니라 ‘나 자신’으로 돌립니다.
2
아첨을 꺼라
Kill the flattery
“내가 틀렸다는 가장 강한 근거”를 일부러 요구합니다.
3
“알아챌 수 있나”
Ask “would I notice”
안 보이던 사각지대를 드러내는 결정적 질문입니다.
사용법 How to use
각 서비스의 최상위 모델로 엽니다 — 클로드라면 페이블(Fable) ·
GPT라면 5.6 솔(Sol). 이 질문들은 판단형이라 모델 지능이 답의
깊이에 그대로 반영돼요.
Open the top-tier model of whichever service you use — Fable on
Claude, 5.6 Sol on GPT. These are judgment questions; model
intelligence shows directly in the depth of the answers.
카드의 프롬프트를 복사해 붙여넣고, 대괄호 안을 당신 상황으로
채웁니다.
Copy a card's prompt, paste it, fill the brackets with your own
situation.
듣기 좋은 답이 아니라 ‘정확한 답’을 받습니다.
You get an accurate answer, not a flattering one.
팁 · Tip — 답이 두루뭉술하면 “더 구체적으로, 아첨 빼고 다시”라고 한
번 더 누르세요. If the answer is vague, push once more: “Be more
specific, cut the flattery.”
10개 한눈에 All 10 at a glance
나를 점검시키는 질문 10 The 10 questions
01
목표 정렬
Goal Alignment
이미 잘하는 걸 다듬느라, 진짜 1순위가 0에 멈춰 있진 않은가.
Are you polishing what you're already good at while the real
priority sits at zero?
쉽게 말하면 — 열심히는 하는데 정작 제일 중요한 일엔 손도 안
댄 경우, AI한테 그걸 골라내게 시키는 거예요.
프롬프트 · KR
여기 내 목표와 지난 2주간 실제로 한 일이 있어.
목표: [내 1순위 목표]
한 일: [지난 2주간 실제로 시간 쓴 일들]
내가 한 일을 두 그룹으로 분류해줘. (1) 1순위 목표를 실제로 전진시킨 것, (2) 이미 잘하는 걸 다듬기만 한 것. 그리고 내가 이 목표의 '현재 숫자'를 모르고 있다면, 오늘 당장 측정해야 할 지표 하나를 알려줘. 듣기 좋은 말은 빼고.
Prompt · EN
Here's my goal and what I actually did over the last two weeks.
Goal: [my #1 goal]
Did: [what I actually spent time on these two weeks]
Sort what I did into two groups: (1) things that genuinely moved my #1 goal, (2) things that just polished what I'm already good at. Then, if I don't know the current number for this goal, name the one metric I should measure today. No flattery.
이럴 때분기 초, 방향이 흐려질 때 ·
Start of a quarter, when priorities blur
02
낡은 자기상
Stale Self-Model
바뀐 건 사실인데, 내 행동은 아직 옛날 버전에 멈춰 있진 않은가.
The facts changed — but is your behavior still running on last
year's snapshot?
쉽게 말하면 — 상황은 바뀌었는데 나만 작년 방식 그대로일 때,
그 지점을 콕 집어달라는 거예요.
프롬프트 · KR
6개월~1년 전의 나와 지금의 나를 비교해줘.
그때: [예전 상황·조건·믿음]
지금: [지금 상황·조건]
내가 아직도 '옛날 버전의 나·고객·시장'을 기준으로 판단하고 있는 지점을 찾아줘. 사실은 이미 바뀌었는데 내 습관·규칙·전제만 안 바뀐 곳을 콕 집어줘.
Prompt · EN
Compare who I was 6–12 months ago with who I am now.
Then: [past situation, constraints, beliefs]
Now: [current situation and constraints]
Find where I'm still deciding based on an outdated picture of myself, my customers, or my market. Pinpoint the spots where the facts already changed but my habits, rules, and assumptions didn't.
이럴 때“원래 이렇게 해왔어”가 반복될 때 ·
When you keep saying “that's how I've always done it”
03
내가 가장 틀린 곳
Where I'm Most Wrong
내 가정이 틀렸다면, 나는 그걸 알아챌 수 있는 구조이긴 한가.
If your biggest assumption is wrong, are you even set up to
notice?
쉽게 말하면 — 내가 틀렸을 때 그걸 알려줄 장치가 없으면, 틀린
채로 1년을 갑니다. 그 사각지대를 찾는 질문이에요.
프롬프트 · KR
내가 지금 [일 / 사업 / 큰 결정]에서 당연하게 믿고 있는 핵심 가정 5개를 뽑아줘.
상황: [내 상황 설명]
각 가정마다 '이게 틀렸다'는 가장 강력한 반론을 세워줘. 그리고 가장 중요한 질문 — 만약 그게 틀렸다면 나는 그걸 알아챌 수 있는 구조인가? '틀려도 못 알아챌' 가정부터 위험 순위를 매겨줘. 아첨 절대 금지, 듣기 싫은 진실 우선.
Prompt · EN
Pull out the 5 core assumptions I take for granted in [my work / business / a big decision].
Context: [describe your situation]
For each, build the strongest possible case that it's wrong. Then the key question: if it were wrong, am I even structured to notice? Rank them by risk, starting with the ones I'd fail to catch even if they're false. No flattery — lead with the truths I don't want to hear.
이럴 때큰 결정 전, 확신이 너무 강할 때 ·
Before a big call, when you feel too certain
04
‘더 낫다’의 정의
Define “Better”
내가 붙잡은 지표가 진짜 목표인가, 재기 쉬워서 붙잡은 허영 지표인가.
Is the metric you track your real goal — or a vanity number
that's just easy to measure?
쉽게 말하면 — 조회수·개수처럼 재기 쉬운 숫자 말고, 진짜
중요한 숫자 하나를 짚어달라는 거예요.
프롬프트 · KR
나는 [내 일 / 사업]에서 '잘하고 있다'를 이렇게 판단해:
[현재 성공을 재는 기준·숫자]
이게 진짜 목표를 재는 지표인지, 아니면 재기 쉬워서 붙잡은 허영 지표(조회수·개수·바쁨)인지 판별해줘. 내 진짜 북극성 지표 1개를 제안하고, 내가 나도 모르게 뒷걸음질치는 걸 잡아낼 체크 항목 3개를 만들어줘.
Prompt · EN
In [my work / business], here's how I judge "doing well":
[the metric or number I currently track]
Tell me whether this measures my real goal, or whether it's a vanity metric I grabbed because it's easy to count (views, volume, busyness). Propose one true north-star metric, and give me three checks that would catch me sliding backward without noticing.
이럴 때바쁜데 성과가 안 느껴질 때 ·
When you're busy but progress doesn't feel real
05
단 하나의 병목
The One Real Bottleneck
이거 하나만 풀면 나머지가 따라 풀리는, 진짜 제약 하나는 무엇인가.
What's the single constraint that, once solved, unblocks
everything else?
쉽게 말하면 — 할 일 100개 중에 '이거 하나 풀면 나머지가 딸려
풀리는' 하나를 찾는 거예요.
프롬프트 · KR
내 [목표]가 막혀 있어. 막힌 지점·증상들이야:
[겉으로 보이는 문제들 나열]
증상을 다 걷어내고, '이거 하나만 풀면 나머지가 따라 풀리는' 진짜 제약 1개를 찾아줘. 그게 정말 뿌리인지 반박 테스트도 해줘(더 깊은 원인은 없나, 이걸 풀면 병목이 어디로 옮겨가나). 마지막에 그 하나에 모든 걸 집중시키는 '이번 주 행동 1개'를 정해줘.
Prompt · EN
My [goal] is stuck. Here are the symptoms and blocked points:
[list the visible problems]
Strip away the symptoms and find the ONE real constraint that, once solved, lets the rest follow. Stress-test that it's actually the root (is there a deeper cause? where would the bottleneck move next?). End with a single action this week that subordinates everything to solving it.
이럴 때할 일은 많은데 안 풀릴 때 ·
When there's a lot to do but nothing's moving
06
반복 결정의 규칙화
Codify Repeated Decisions
매번 새로 고민하는 것 중, 사실은 규칙 한 줄로 끝낼 수 있는 건
무엇인가.
Which decisions you re-agonize over could just become a one-line
rule?
쉽게 말하면 — 매번 똑같이 고민하는 걸 '규칙 한 줄'로 만들어서
다시는 고민 안 하게 하는 거예요.
프롬프트 · KR
내가 최근 반복해서 내린 결정들이야:
[반복되는 결정·판단 상황들]
이 중 매번 새로 고민할 필요 없이 '고정 규칙'으로 만들어야 할 것과, 매번 판단이 필요해서 규칙화하면 안 되는 것을 구분해줘. 규칙화할 것은 내가 그대로 쓸 수 있는 '한 줄 규칙 문장'으로 써줘.
Prompt · EN
Here are decisions I've made over and over lately:
[the recurring decisions or judgment calls]
Separate the ones I should turn into a fixed rule from the ones that genuinely need fresh judgment each time (and shouldn't be ruled). For the ones to codify, write each as a one-line rule I can use as-is.
이럴 때같은 고민을 자꾸 반복할 때 ·
When you keep re-having the same debate
07
무엇을 맡길까
What to Delegate
매번 확인하느라 느린 곳과, 확인 없이 저질러서 위험한 곳은 각각
어디인가.
Where are you too slow (over-checking) and where are you too
risky (acting without a check)?
쉽게 말하면 — AI한테 뭘 알아서 시키고, 뭘 내가 꼭 쥐고 있어야
하는지 선을 그어주는 거예요.
프롬프트 · KR
내가 [AI / 팀원 / 나 자신]에게 맡기는 일들이야:
[일 목록]
각각을 세 단계로 나눠줘. (1) 물어보지 말고 알아서 해도 되는 것, (2) 실행 전 반드시 확인받아야 하는 것, (3) 아예 손대면 안 되는 것. 그리고 내가 쓸데없이 매번 확인하느라 느린 곳과, 확인 없이 저질러서 위험한 곳을 각각 짚어줘.
Prompt · EN
Here are the things I delegate to [AI / a teammate / myself]:
[list of tasks]
Sort each into three tiers: (1) just do it, no need to ask, (2) must get a check before acting, (3) don't touch at all. Then point out where I'm needlessly slow from over-checking, and where I'm risky from acting without a check.
이럴 때AI·팀에 위임을 늘리기 전 ·
Before you hand more over to AI or a team
08
10배 vs 소멸
What 10×'s, What Dies
AI가 더 세지면, 내 기술 중 무엇이 10배가 되고 무엇이 그냥 죽는가.
As AI gets stronger, which of your skills get 10× — and which
just die?
쉽게 말하면 — AI가 세지면 내 일 중에 뭐가 10배가 되고 뭐가
사라지는지, 그래서 시간을 어디로 옮길지 알려줘요.
프롬프트 · KR
내 일의 구성 요소야:
[내가 하는 일·가진 기술·자산 목록]
앞으로 1~2년 AI가 더 강해질 때, 각 항목을 '10배로 증폭될 것 / 그냥 죽을 것 / 섞임'으로 분류해줘. 기준은 이거야 — '손으로 만드는 시간'은 죽고, '무엇을 만들지 정하고·검증하고·관계로 파는 것'은 산다. 그래서 내 시간을 어디에서 어디로 옮겨야 하는지 알려줘.
Prompt · EN
Here are the parts of my work:
[what I do, my skills, my assets]
As AI gets stronger over the next 1–2 years, sort each into: gets 10×'d / just dies / mixed. Use this rule — hands-on "making time" dies, while "deciding what to make, verifying it, and selling through relationships" survives. Then tell me where to move my time, from what to what.
이럴 때커리어·스킬 투자를 정할 때 ·
When deciding where to invest your skills next
09
나 없으면
Bus Factor
내가 30일간 사라지면 무엇이 무너지나 — 오직 나만 아는 건 무엇인가.
If you vanished for 30 days, what breaks — and what does only you
know?
쉽게 말하면 — 나 혼자 아는 것·나만 할 수 있는 걸 찾아서, 지금
미리 넘기거나 적어두게 하는 거예요.
프롬프트 · KR
만약 내가 내일부터 30일간 완전히 사라진다면, 내 [일 / 사업 / 삶]에서 무엇이 무너질지 목록으로 만들어줘.
지금 상황: [내가 혼자 도맡고 있는 것들]
오직 나만 아는 것·나만 할 수 있는 것을 찾아줘. 각각에 대해 지금 당장 '남기거나(문서화) · 자동화하거나 · 넘길(위임)' 수 있는 최소 조치를 하나씩 제안해줘. 가장 위험한 것부터.
Prompt · EN
If I vanished completely for 30 days starting tomorrow, list what would break in my [work / business / life].
Current setup: [what you single-handedly carry]
Find what only I know and only I can do. For each, propose the smallest step to either document it, automate it, or hand it off — starting with the most dangerous.
이럴 때혼자 다 짊어지고 있다고 느낄 때 ·
When it all rests on you alone
10
빅 픽처
Big Picture (Ikigai)
내가 지금 쓰는 시간이, 내 진짜 레버리지의 교집합 안에 있는가.
Is the time you spend today inside the sweet spot where your real
leverage lives?
쉽게 말하면 — 바쁜 걸 잠깐 멈추고, 방향 자체가 맞는지 한 번
크게 보는 질문이에요.
프롬프트 · KR
나에 대한 정보야.
좋아하는 것: [ ]
잘하는 것: [ ]
세상이 필요로 하는 것: [ ]
돈이 되는 것: [ ]
이 네 개가 겹치는 교집합을 찾아줘. 지금 내가 시간을 쓰는 일 중 이 교집합 밖에 있는 걸 지적하고, 교집합 안으로 이동하는 구체적인 첫 수 하나를 제안해줘.
Prompt · EN
Here's what I know about myself.
Love doing: [ ]
Good at: [ ]
The world needs: [ ]
Pays well: [ ]
Find the overlap of these four. Point out what I currently spend time on that sits outside that overlap, and propose one concrete first move to shift into it.
이럴 때방향 자체를 다시 잡고 싶을 때 ·
When you want to reset your direction, not just tasks
한 가지 더 — 답을 파일로 남기세요
One more — save the answers as files
좋은 답을 받는 것과 그 답이 남는 것은 다릅니다. 채팅창을 닫는 순간 그
답은 사라져요.
Getting a good answer and keeping it are two different things. Close
the chat, and the answer is gone.
저는 이 질문들로 나온 결론을 전부 실제 규칙 파일로 만들어 박았습니다.
가격 기준·반복 결정·마감 알림 같은 것들이요. 여러분 버전은 더
간단합니다. 답에서 나온 규칙과 기준을 메모 하나(“나의 운영 규칙”
같은)로 정리한 다음, AI가 항상 먼저 읽는 곳에 한 번 넣어두세요 —
ChatGPT는 메모리나 맞춤 지시, 클로드는 프로젝트, 제미나이는 저장 정보.
그럼 매 대화에 자동으로 적용됩니다. 이런 기능이 없는 도구를 쓰거나
오늘따라 확실하게 적용돼야 하는 날엔 그 메모를 대화 맨 앞에 직접
붙여넣으면 돼요. 자동 메모리는 가끔 빼먹지만 붙여넣기는 빼먹지
않으니까요.
I turned every conclusion from these questions into actual rule
files — pricing standards, recurring decisions, deadline alerts.
Your version is simpler: gather the rules and standards from the
answers into one note (call it “my operating rules”), then put it
once where your AI always reads first — Memory or custom
instructions on ChatGPT, a Project on Claude, saved info on Gemini.
It then applies to every conversation automatically. If your tool
has no such feature, or it's a day when the rules absolutely must
apply, paste the note at the top of the chat — auto-memory sometimes
skips things; pasting never does.
그래서 순서가 중요합니다. ① 점검은 최상위 모델로 한 번 — 여기에만 비싼
지능을 쓰세요. ② 나온 규칙을 AI의 기억 공간에 저장. ③ 일상 작업은
가벼운 모델로 그 바닥 위에서. 비싼 모델은 점검에, 가벼운 모델은 실행에
쓰는 구조예요.
So the order matters. ① Run the audit once on the top-tier model —
spend the expensive intelligence only here. ② Save the resulting
rules into your AI's memory. ③ Do everyday work on a lighter model,
on top of that floor. Expensive model for the check-up, lighter
model for the daily work.
쉽게 말하면 — 모델은 바뀌어도 파일은 남습니다. 이렇게 해두면
나중에 어떤 AI를 쓰든 — 더 가벼운 모델로 바꿔도 — 업그레이드된 바닥
위에서 시작해요. 물론 그 자리에서 내리는 새 판단의 깊이는 모델마다
다르지만, 바닥은 한 번 올려두면 내려가지 않습니다.
Models change; files stay. Do this and whichever AI you use later —
even a lighter model — starts on the upgraded floor. The depth of
new judgment still varies by model, but the floor, once raised,
never drops.
마지막 한 줄 One last line
10개를 다 쓸 필요는 없습니다. 지금 가장 찔리는 카드 하나를 골라,
대괄호를 채우고, 붙여넣으세요. AI가 당신에게 되묻는 순간, 결과물이
아니라 당신이 바뀝니다.
You don't need all 10. Pick the one card that stings most right now,
fill the brackets, paste it. The moment AI turns the question back
on you, it's not the output that changes — it's you.
AI 활용, 혼자 말고 같이
이런 질문과 실전 AI 활용법을 나누는 오픈채팅방이 열려 있어요. 편하게
들어와서 구경하세요. An open chat where we share prompts and
real-world AI know-how — come take a look.