Part 1 · Bari Picture Se Start

1. AI Access, AI Fluency Nahi Hai

Ek powerful AI tak access hona, use achi tarah use karna aane ke barabar nahi. Almost sab ke paas same models available hain. Same plan pe same assistant use karne wale 2 log bilkul alag results paate hain. Farq tool mein nahi tha, farq is baat mein tha ke unhone use kaise use kiya.

AI Fluency Ka Matlab: AI Ke Sath Aise Kaam Karna Jo

Jo cheez zaroori NAHI haiwo ye hai: large language models kaise train hoti hain ye samajhna, transformer architecture ki deep knowledge, ya “magic prompts” ka koi collection. Foundation bas itni hai: AI ke around achhe human decisions lena seekhna. Sequence yehi hai: Delegation, phir Description, phir Discernment, phir Diligence — ya simple words mein: decide karo, explain karo, check karo, zimmedari lo.

Pichle Chapter Se 3 Baatein Yaad Rakho

• Plausible, correct ke barabar nahi hai — AI aisa confident jawab de sakta hai jo galat ho (hallucination)
• Output badalta rehta hai, same request alag waqt pe alag jawab de sakti hai
• AI sirf available information pe kaam karta hai, jo info missing hai wahan guess ho sakta hai

2. AI Ke Sath Kaam Karne Ke 3 Tareeke: Automation, Augmentation, Agency

4Ds seekhne se pehle, ek aur foundational idea zaroori hai. Log AI ke sath 3 broad modes mein kaam karte hain, aur inka farq mainly is baat se hota hai ke AI ko agla step khud decide karne ki kitni azaadi di gayi hai.

Automation: “Ye Task Karo”

Automation mein aap AI ko exactly batate ho kaunsa task karna hai, jaise “is report ko 5 bullets mein summarize karo”, “is email ka Urdu translation karo”, ya “invoice se number, date, total nikaal do.” Yahan aap ek script writer ho. Ye tab sabse achha kaam karta hai jab task clear aur repeatable ho.

Augmentation: “Mere Sath Socho”

Augmentation mein aap aur AI mil kar kaam karte ho — jaise business idea pe brainstorm karna, software architecture review karna, lesson plan behtar banana, do strategies compare karna, ya koi aisa sawal explore karna jiska jawab aapko khud abhi pata nahi. Yahan AI ek thinking partner ki tarah kaam karta hai, sirf instructions execute nahi karta. Aap aur AI ke beech kai turns back and forth hote hain: aap poochte ho, wo jawab deta hai, aap challenge karte ho, wo revise karta hai.

Agency: “Ye Goal Mere Liye Achieve Karo”

Agency mein aap AI ko ek goal aur boundaries de dete ho, phir usay kai steps khud decide karne dete ho. “Ye 5 emails parh ke summary do” kehne ki jagah, aap kuch aisa bolte ho:

"Keep my inbox manageable. Reply to routine messages, flag important ones,
and ask me before doing anything you are unsure about."

Ab AI ko khud decide karna parta hai ke kya routine hai, kya important hai, aur kab poochna hai. Yahan aap script writer se director ban jate ho. 2 words yahan bohot weight rakhte hain. Pehla, future — matlab aap us waqt present nahi ho, aap ne Monday ko sab set kar diya aur Thursday ka kaam khud ho raha hai jab aap so rahe ho. Doosra, for others — matlab jis insaan ki AI service kar rahi hai, wo khud aap na bhi ho. Automation aur augmentation mein aap chair pe baithe rehte ho, lekin agency mein aap chair se uth jate ho. Mode 2 ki har mushkil isi ek fact se aati hai: aap har decision supervise nahi kar sakte, isliye judgment pehle se built-in honi chahiye.

Koi mode dusre se automatically behtar nahi hota. Ek achha AI user wahi mode pick karta hai jo task ko chahiye, aur ek project mein teenon mode use ho sakte hain. Mode 1 zyada tar automation aur augmentation use karta hai, jabke Mode 2 agency ko systematic bana deta hai. Ek Digital FTE sirf itna nahi ke “AI kaam kar rahi hai”, balke ye ek job definition, System of Record, permissions, rules, aur governance ke andar reh kar kaam karta hai.

Part 2 · 4 Competencies

3. Delegation: Decide Karo Kaun Kya Karega

Beginners ki sabse common ghalti pehle prompt se bhi pehle ho jati hai. Log AI assistant khol kar seedha type karna shuru kar dete hain, jabke unhone abhi ye decide hi nahi kiya hota ke actually achieve kya karna hai, achha result kaisa dikhega, kaunsa hissa AI kare, kaunsa khud karein, aur kaunse decisions kabhi AI ko na diye jayen. Yehi Delegation ka problem hai.

3.1 Problem Awareness

AI se kuch bhi poochne se pehle, khud se ye poocho:

• Goal kya hai?
• Ye kiske liye hai?
• Success kaisa dikhega?
• Kya galat ho sakta hai?
• Human judgment kahan zaroori hai?

Ek example dekho: ek beginner AI se kehta hai, “mujhe ek invoice-chasing agent bana do.” AI kuch na kuch bana to dega, lekin kai hard sawal abhi bhi unanswered rahenge — kaunse customers ko contact karein, invoice kitne din late hone pe follow-up bheja jaye, kaunsa tone use ho, kitni amount pe human ki approval chahiye, agar customer invoice pe dispute kare to kya kiya jaye, agent kaunsa accounting system parh sakta hai, aur ye ke agent sirf message ka draft banaye ya khud bhej bhi sake. Ye business questions hain, prompting questions nahi. AI aapki business policy khud se decide nahi kar sakta jab tak aap jaan-boojh kar usay ye authority na dein — aur zyada tar cases mein aapko dena bhi nahi chahiye.

3.2 Platform Awareness

Har AI system har task mein equally achha nahi hota. Mushkil, multi-step problems ke liye reasoning model pick kar sakte ho, current info ke liye search-enabled assistant, software banane ke liye coding agent, aur un tasks ke liye jinme tools aur multiple steps shamil hon, agent-capable system. Ek achi habit ye hai ke khud se poocho, “kya ye tool is job ke liye sahi hai?” Alag alag systems try karo, unke results compare karo, aur note karte raho ke kya kaam kar raha hai.

3.3 Task Delegation

Jab problem aur platform, dono samajh mein aa jayen, to kaam ko soch samajh kar parts mein divide karo. Ek example, ek course banate waqt:

Agent Factory mein Delegation ek engineering step ban jati hai. Problem Awareness ek specification ban jati hai — yani goal, constraints, risks, aur “definition of done”. Task Delegation Digital FTE ki boundary ban jati hai, jo decide karti hai AI kya kar sakta hai, humans kya apne paas rakhte hain, aur kya escalate hona chahiye. Aur ek vertical System of Record in decisions ko durable aur inspectable banata hai.

4. Description: AI Ko Wo Do Jo Usay Chahiye

Shuru wale colleague ko yaad karo, uski outline isliye galat thi kyunki usay zaroori information nahi di gayi thi. AI ko bhi yehi problem hoti hai, magar zyada severely. Wo aapka mind read nahi kar sakta. Agar aap koi zaroori cheez chhor dein, to wo ek reasonable guess laga leta hai — aur ye reasonable guess bhi galat ho sakti hai.

4.1 Product Description: Result Define Karo

Ye sawal poochti hai: “mujhe wapis exactly kya chahiye?”

Vague:

Summarize this report.

Clear:

Summarize this quarterly financial report for senior executives who have
ten minutes to read. Focus on revenue trends, major risks, and recommended
actions. Use short bullet points and keep it to one page. Highlight any
figure that changed significantly from last quarter. Avoid unnecessary
accounting jargon.

Doosra request zyada intelligent nahi, zyada complete hai. Complete hona aksar clever wording se zyada matter karta hai.

4.2 Process Description: Approach Define Karo

Ye sawal poochti hai: “AI ko kaam kaise karna chahiye?” Example:

Review this code for correctness first, security second, and style last.
Do not spend time on naming issues until you have checked whether the
code actually works.

Process description sabse zyada tab matter karti hai jab kaam ke kai stages hon. Ek example, 3 vendors — A, B, aur C — ke proposals se ek recommendation banana:

1. 1

   Extract

   Har proposal se same facts ek table mein nikalo: price, contract length, exit terms, support hours. Check: 3-4 cells ko source se confirm karo.

2. 2

   Compare

   Sirf table use kar ke vendors compare karo. Check: har farq actually table mein hai?

3. 3

   Score

   Har vendor ko score do, jo cheez aapke liye zyada matter karti hai usay zyada weight do. Check: scores aapki criteria follow karte hain, ya AI ne khud koi criterion add kar diya?

4. 4

   Draft

   Recommendation likho. Check: ye sirf wahi claim karta hai jo steps 1-3 support karte hain?

4.3 Performance Description: Behavior Define Karo

Ye sawal poochti hai: “ye AI kaise behave kare, aur kis ke sath?” Example:

Challenge my assumptions when they are weak. Flag uncertainty. Do not
agree with me just to be polite. If my argument is stronger, explain why.
If yours is stronger, hold your position and explain it.

Isse AI ek polite jawab machine se behtar thinking partner ban jata hai. Iske 2 versions hain. Chhota version, chat window mein, sirf aapke liye. Bara version, ek deployed agent mein — jaise ek tutoring agent ka rule “student ke attempt karne se pehle answer mat do.” Ye same sentence hai bas stakes badal gaye: ek baar likha jata hai, aur hazaar baar un logon pe apply hota hai jinse aap kabhi mile hi nahi. Chat mein ek weak performance description 10 minute annoy karti hai, lekin ek deployed agent mein wahi khud product ban jati hai.

Description, Prompting Se Bigger Hai

Book ka agla chapter, “AI Prompting in 2026”, practical techniques sikhata hai — jaise examples dena, constraints specify karna, tasks ko chhote parts mein todna, aur roles define karna. Ye chapter is se bhi bigger idea sikhata hai: Description.

Kai teams ek prompt template use karti hain jisme role, context, task, constraints, aur output format ke liye jagah hoti hai. Task, constraints, aur output format, actually product description hi hain. Context wahi cheez hai jise agla section context engineering kehta hai. Agar kaam kaise aage badhega iske liye ek line, aur AI aapke sath kaise behave karega iske liye ek line add kar do, to ye template teenon parts cover kar leta hai.

Prompt Engineering Se Context Engineering Tak

Prompt engineering ye sawal poochti hai: “ye message kaise likhun?” Context engineering ek bigger sawal poochti hai:“AI ke succeed karne ke liye kya available hona chahiye?”

• Documents
• Examples
• Memory
• Conversation history
• Policies
• Tools
• Database records
• Definitions
• Instructions

Ek agent ke liye ye bohot matter karta hai: ek beautiful prompt bhi agent ko nahi bacha sakta agar uske paas galat data ho, rules missing hon, weak examples hon, ya zaroori tools tak access na ho. Agent Factory mein: system prompt ek performance description ko persistent banata hai, ek SKILL.md ek process description ko reusable banata hai, aur ek System of Record domain knowledge, rules, definitions, aur governance rakh sakta hai.

5. Discernment: Confidence Ko Correctness Mat Samjho

AI aksar confident sound karta hai. Ye readability ke liye useful hai, lekin trust ke liye risky hai. Ek galat AI answer usually kisi warning label ke sath nahi aata, wo polished, detailed, aur certain dikh sakta hai.

Fake Answer Bhi Success Jaisi Dikh Sakti Hai

Ek hallucinated answer, correct answer se almost identical dikh sakta hai. Fluent writing koi proof nahi, citation jaisi link koi proof nahi, confident tone bhi koi proof nahi. 4 signs jo aksar fake answer show karte hain:

Discernment, Description Ko Mirror Karti Hai

Isi ke bhi 3 parts hain, aur ye 3 parts unhi 3 cheezon ki taraf point karte hain:

5.1 Product Discernment

• Kya ye factually correct hai?
• Kya har zaroori requirement follow hui?
• Kuch missing to nahi?
• Kya ye internally consistent hai?
• Kya ek expert isay credible samjhega?
• Kya main is par apna naam laga sakta hoon?

Ek zaroori baat:yahan aapki apni domain knowledge bohot valuable ban jati hai. Ek accountant weak assumptions pakar leta hai, ek programmer subtle bugs pakar leta hai, aur ek teacher wo explanations pehchan leta hai jo beginners ke liye confusing hon. “AI expert work ki speed barha sakta hai, lekin ye expertise ki zaroorat khatam nahi karta.”

Result ke peeche ki reasoning ko bhi judge karo, kyunki AI weak reasons ke sath bhi correct answer tak pahunch sakta hai, aur ek galat assumption pe khara correct answer zyada der correct nahi rehta. Useful ye hai ke AI se ye dikhwao: assumptions, evidence, decision criteria, calculations, intermediate results, aur alternative interpretations. Inhe ek review ke liye justification samjho, hidden reasoning ka literal transcript nahi.

Before recommending one option, list your assumptions, the evidence
supporting them, and the criteria you are using to decide. Then give
the recommendation.

Document-based answers ke liye ye bhi add karo:

Answer only from the three proposals I attached, not from anything
you know about these vendors. If a proposal does not state its exit
terms, say so instead of guessing. For every term you report, quote
the proposal's section heading and the sentence it came from.

Ye 3 kaam karta hai. Pehla, sirf wahi cheez use karta hai jo attach ki gayi ho, taake training knowledge se gaps na bharein, jahan se imaginary features aati hain. Doosra, ye “isme ye likha nahi” kehne ki permission deta hai. Teesra, ye aapko ek minute mein checkable cheez deta hai, teenon proposals dobara parhne ki jagah. Koi bhi cheez answer khud parhne ki jagah nahi le sakti, ye sirf reading fast bana deti hai aur fake answer pakarna aasan kar deti hai.

5.2 Process Discernment

Kabhi kabhi answer theek hota hai lekin AI ke sath kaam karne ka tareeka theek nahi hota. Ye wo sawal hai jo log almost kabhi nahi poochte, kyunki output acceptable laga aur session successful feel hua. Khud se ye poocho:

• Kya AI mere feedback ke hisab se adapt kar raha hai, ya wapis purani cheez pe drift ho raha hai?
• Kya do baar correct karne ke baad bhi wahi ghalti repeat kar raha hai?
• Kya wo itna agreeable ho gaya hai ke ab useless ban gaya hai?
• Kya main har turn wahi formatting problem khud fix kar raha hoon?
• Kya main iska draft khud likhne se zyada heavily edit kar raha hoon?

Jab ye process kaam nahi kar raha, teen escalating moves hain, aur teenon fluency hain: performance description badlo, tool badlo, ya task wapis khud apne paas le lo. Sirf teesra defeat jaisa feel hota hai, lekin usually aisa hota nahi.

5.3 Performance Discernment

Ye tab exist karti hai jab aap agency use kar chuke ho, aur book ke end mein isi ki sabse zyada care ki jati hai. Ye ye sawal poochti hai ke kya AI ka independent, user-facing behavior logon ke liye achhe outcomes deta hai. Ye is baat se different hai ke koi ek output correct tha ya nahi. Example: ek AI tutor har sawal ka correct jawab de sakta hai, aur phir bhi bad tutor ho sakta hai agar wo student ke hesitate karte hi solution de deta hai, jisse koi kuch nahi seekhta. Isi tarah ek support agent tickets ko fast resolve kar sakta hai, aur phir bhi bad ho sakta hai agar wo conversation us waqt band kar de jab customer ne usay abhi complete na samjha ho. Ye single chats mein nahi, aggregate mein show hota hai: users iske baad kya karte hain, wo kis baat pe complain karte hain, aur kaunse cases baar baar chupke se wahi galat hote hain. Aap thousands conversations khud check nahi kar sakte, isliye kuch aisa banana parta hai jo aapki jagah in par nazar rakhe.

Description–Discernment Loop

Jab Discernment koi problem flag kare, to usually iska fix behtar description hota hai. Kabhi kabhi ye wapis Delegation tak le jata hai — ya to galat tool ki wajah se, galat split ki wajah se, ya phir galat approach ki wajah se. “Professional AI collaboration iteration se converge hoti hai, ye kabhi ek hi shot mein perfect nahi hoti.”

Ye Normal Hai

AI ke sath achha kaam usually iterative hota hai. Pehla response aksar ek draft hota hai, finish line nahi. Feedback dete waqt is pattern ka use karo: problem batao, phir ye ke wo kyun matter karta hai, phir direction do.

Weak feedback:

Wrong. Try again.

Better feedback:

The second section assumes enterprise customers. Our audience is solo
founders, so the advice is too expensive. Rewrite that section for a
one-person business with a limited budget.

Ek Achhe Draft Ko Doosri Try Ki Zaroorat Kyun Parti Hai

Kyunki wo galat reader ke liye likha gaya tha. Ek example, ek quarterly report ki finding, board aur support team, dono ke liye alag alag likho:

Facts same, model same, results dono different. In mein se koi bhi doosre reader ke liye kaam nahi karta.

Har Review Ka Ek Ending Hota Hai

• Kaam bhej diya jata hai, jaise cycling club wala email send ho gaya
• Ye feedback ke sath wapis chala jata hai
• Aap khud kaam apne haath mein le lete ho, kyunki fix ke liye sirf aap jaante ho

Kabhi kabhi behtar description bhi kaafi nahi hoti, aur Discernment dikha deti hai ke original Delegation decision hi galat tha — shayad galat tool choose hua, shayad AI ko wo hissa kabhi milna hi nahi chahiye tha, ya shayad ye kaam ek human expert hi maangta hai. Ye bhi fluency hai. Agent Factory mein Discernment evaluation engineeringban jati hai — yani aapka personal sawal “kya ye kaafi achha hai?” eval suites, production checks, monitoring, sampling, aur release gates ki shape le leta hai. “Aap wo judgment automate nahi kar sakte jo aapne khud kabhi seekha hi nahi.”

6. Diligence: AI Ko Responsibly Use Karna

Pehli 3 Ds behtar results lane mein help karti hain. Diligence ek alag sawal poochti hai:

“Kya mujhe AI ko is tarah use karna bhi chahiye?”

6.1 Creation Diligence

Koi bhi information share karne se pehle ye poocho:

• Kya isme personal data hai?
• Kya isme company ki confidential information hai?
• Kya mujhe ye info is tool mein daalne ki permission hai?
• Ye data kaun access ya retain kar sakta hai?
• Kya ye service meri organization se approved hai?
• Koi legal, contractual, ya professional restrictions hain?

Aasan raasta hamesha responsible raasta nahi hota. Iska fix aksar ye nahi hota ke task chhor diya jaye, balke ye hota hai ke data ko stripkar diya jaye. Teacher wale example mein, wo naam aur student ID hata sakta tha, sirf grade range aur ek behavior rakh sakta tha, aur usi se feedback draft kar sakta tha. Principle ye hai: “AI ko pattern chahiye, person nahi.” Isay redaction kehte hain.

6.2 Transparency Diligence

Har AI-assisted task ko public announcement ki zaroorat nahi hoti. Lekin jab AI dusre logon ko significantly affect kare, to disclosure matter karti hai: academic work, hiring decisions, customer communications, medical ya financial advice, professional reports, ya aisa content jo original human work ki tarah present kiya jaye. Exact rules context, organization, law, aur professional standard pe depend karti hain.

6.3 Deployment Diligence

AI-assisted kaam publish, send, execute, ya kisi decision mein use hone se pehle check karo. Kitni checking chahiye, ye transparency ka principle decide karta hai — jitne zyada log affect hote hain, utni zyada checking chahiye: khud ke liye likha note ek nazar leta hai, welcome email full read leta hai, aur regulator ko jane wali report ek second reviewer se guzarti hai.

• Facts verify karo
• Confirm karo ke sources actually exist karte hain
• Calculations check karo
• Bias ya unfair outcomes ka review karo
• Permissions aur rights confirm karo
• Organization ki policy follow karo
• High-impact actions ke liye human approval lo

The Numbers Rule

Ek Powerful Last Question

“Kya main confidently is par apna naam laga sakta hoon?”

Agar jawab “na” hai, to kaam ready nahi hai.

Kabhi kabhi case clear galat nahi hota, sirf unclear hota hai. Example: ek AI-ranked job applicant shortlist reasonable lagti hai, lekin pata nahi chalta ke wo silently do universities ke graduates ko favor kar rahi hai ya nahi. Decide karne se pehle ye 4 sawal poocho:

• Is result se kaun affect hota hai, un logon samet jo ise kabhi dekhenge bhi nahi?
• Unke liye kya galat ho sakta hai, aur kya wo bata bhi payenge?
• Yahan fair outcome kaisa dikhega?
• Kya kisi cheez ka disclose hona chahiye, aur kis ko?

Agar aap chaaron ka jawab de sakte ho, to decide karo aur likh lo. Agar nahi, to guess kar ke bhejne ki jagah, us decision ke owner tak baat pahunchao. Guessing ek unclear case ko aapki apni ghalti bana deta hai.

“AI kaam khud kar sakta hai, lekin responsibility khud nahi le sakta.”

Agar ek AI-assisted system koi harmful decision le leta hai, to us system ko chalane wali organization phir bhi responsible rehti hai. Agar coding assistant koi vulnerability introduce kar de aur engineer usay ship kar de, to engineer aur organization, dono result ke owner rehte hain. Agent Factory isliye governance ko sabse pehle rakhta hai: Creation diligence data rules, access control, aur approved-tool policy ban jati hai. Transparency diligence disclosure aur user experience design ban jati hai. Deployment diligence evaluation gates, audit logs, monitoring, aur human review ban jati hai.

Part 3 · Chaaron Ko Ek Sath Milana

7. 4Ds Ek Practical Operating Loop Ki Tarah

Delegation decide karti hai AI kaam mein aaye ya nahi, aur kya own kare. Description AI ko goal, context, process, aur behavior deti hai. Discernment result check karti hai aur agla round behtar banati hai. Diligence poore process ko zimmedari se cover karti hai.

Example: Ek Bookkeeping Digital FTE

Ayesha Lahore mein ek Forward Deployed Engineer hai, jo Karachi ki ek small accounting practice ke liye bookkeeping Digital FTE bana rahi hai. Pehla kaam jo automate karna hai, wo hai monthly bank reconciliation.

1. 1

   Step 1 · Delegation

   Ayesha ye nahi poochti ke "reconciliation agent bana do." Wo accounting partners ke sath baith kar poora kaam map karti hai: agent bank transactions ko ledger entries se match kar sakta hai, unmatched items flag kar sakta hai, aur report draft kar sakta hai. Humans har journal adjustment approve karenge, har write-off ka decision apne paas rakhenge, tax se related koi bhi cheez accountant ke paas rahegi, aur high-value unmatched items ek named person tak escalate hongi.

2. 2

   Step 2 · Description

   System ko chart of accounts, matching rules, purani reconciliations ki examples, partners ka report format, escalation rules, aur duplicate/stale cheque ki definitions di jati hain. Rule ye hai: agent kabhi khud journal entry post nahi karega, aur kabhi client se directly contact nahi karega.

3. 3

   Step 3 · Discernment

   Ayesha ye assume nahi karti ke agent sirf isliye theek kaam kar raha hai kyunki demo achi lagi. Wo agent ko purani, trusted reconciliations ke against test karti hai: kitne matches correct hain, kitne galat matches slip karte hain, sahi cases escalate hote hain ya nahi, kya agent zaroorat se zyada escalate kar raha hai, aur kya performance waqt ke sath change ho rahi hai. Ek accountant kuch "successful" matches ka bhi review karta hai, sirf failures ka nahi, kyunki koi system silently fail ho kar bhi safe dikh sakta hai.

4. 4

   Step 4 · Diligence

   Client ka financial data approved infrastructure ke andar rehta hai, agent ke actions log hote hain, jahan zaroori ho wahan clients ko bataya jata hai ke reconciliation AI-assisted hai, aur phir bhi ek human partner reconciliation sign karta hai aur final result ka zimmedar rehta hai.

Yehi 4D loop practice mein hai. Ek personal skill ek system property ban gayi.

Chart: Chat Skill Se Factory System Tak

“Agent Factory AI fluency replace nahi karta. Wo use industrialize karta hai.”

10-80-10 Rule Se Connection

8. Beginners Ki 4 Common Mistakes

Daily Checklist

Har chhote task ke liye ise paperwork banane ki zaroorat nahi, maqsad bas itna hai ke ye 4 sawal apne aap zehan mein aa jayen.

Practice Se Pehle Ek Chhota Recap

AI fluency prompts ratta lagane ki ability nahi hai. Ye AI ke sath effectively, efficiently, ethically, aur safely kaam karne ki ability hai. Iske 3 modes hain: Automation (AI defined task karta hai), Augmentation (aap aur AI sath sochte hain), aur Agency (AI aapke set kiye goal ki taraf khud kaam karta hai, aksar un logon ke liye jo aap khud nahi hain).

“Decide karo AI kya kare. Kaam ko clearly describe karo. Jo wapis aaye usay check karo. Aage jo bhi ho uska zimmedar bano.”

Ab Khud Try Karo: 6 Prompts

Sirf parhna kaafi nahi, ek AI assistant khol kar ye exercises khud try karo. Sab ek hi sitting mein complete karna zaroori nahi.

1. 1

   1. Kisi Real Task Ke Liye 4D Plan Banao

   Koi real task chuno aur AI se poocho ke wo ek ek kar ke Delegation, Description, Discernment, Diligence pe sawal poochhe, jo apply na ho wo skip kar de, aur end mein ek chhoti table de. Dhyan do: plan zyada tar aapke apne answers se banta hai, AI ke nahi, kyunki Delegation aur Diligence sirf aap khud decide kar sakte ho.

2. 2

   2. Ek Jani-Pehchani Topic Pe Discernment Try Karo

   Ek aisi topic pe baat karo jismein aapko actually experience ho, aur AI se kaho ke wo ek knowledgeable colleague ki tarah baat kare, teacher ki tarah nahi. Dhyan do: jab domain aapki khud ki ho to Discernment kitni cheap feel hoti hai, aap bina effort ke galat claim pakar lete ho.

3. 3

   3. Non-Expert Hone Ka Feel Karo

   Ek aisi topic chuno jisme aapko kuch bhi pata na ho, aur AI se kaho ke wo ek beginner ke liye explain kare, aur end mein wo claims batae jo aapko verify karni chahiyen. Dhyan do: same quality output kitna different feel hota hai jab check karne ko kuch bhi na ho — yehi feeling har us user ki hai jo aapka banaya agent use karega.

4. 4

   4. Ek Performance Description Likho

   Session ke shuru mein hi AI ko bata do ke wo weak assumptions ko challenge kare, uncertainty flag kare, aur sirf politeness ke liye agree na kare. Dhyan do: farq kitni jaldi visible hoti hai, aur agar naya chat khol kar ye dobara na batao, to ye kitni jaldi fade ho jati hai.

5. 5

   5. Recommendation Se Pehle Uski Justification Check Karo

   Koi real decision AI ko do aur usay assumptions, evidence, criteria, aur uncertainties list karne ko kaho, phir recommendation mangwao. Dhyan do: koi bhi assumption jo aap bina likhe silently accept kar lete, wahi sabse zyada check karne layak hai.

6. 6

   6. Ek Chhota Project Poore 4D Loop Se Guzaro

   Ek hour mein complete hone wala project chuno: Delegation se shuru karo, har AI-owned task se pehle description poocho, har zaroori output ke baad ruk kar evaluate karo, aur end mein facts, sensitive data, disclosure, aur approvals ka Diligence check karo.

Is Chapter Ke Naye Terms

Exam ke liye ye poori glossary yaad rakho, koi bhi term skip mat karo:

Self-Test

Khud Se Poocho

Pehle khud answer do, phir sawal pe click kar ke answer check karo. Agar 8+ correct hon, to aap agle chapter ke liye ready ho.