Prompt

1️⃣ Zero-Shot Prompt

What it is:No examples, just instructions.

When to use:

• Simple, well-defined tasks

• Fast responses

• When model already “knows” the domain

Example:

“Summarize this loan agreement in 5 bullet points.”

Enterprise use:

• Email drafting

• Basic summaries

• Simple Q&A bots

⚠️ Risk: Output can vary

• Not reliable for compliance-critical use

2️⃣ One-Shot Prompt

What it is:Instruction + 1 example

When to use:

• When output format matters

• Light guidance improves accuracy

Example:

“Classify transaction risk.Example: ‘₹5,00,000 international transfer’ → High RiskNow classify: ‘₹10,000 domestic UPI transfer’”

Enterprise use:

• Transaction categorization

• Ticket classification

3️⃣ Few-Shot Prompt

What it is:Instruction + multiple examples

When to use:

• Complex classification

• Domain-specific behavior

• Consistent output needed

Example:Provide 3–5 labeled fraud examples, then ask model to classify a new one.

Classify loan application risk.

Example 1:

Income: ₹20L, CIBIL: 820 → Low Risk

Example 2:

Income: ₹4L, CIBIL: 580 → High Risk

Example 3:

Income: ₹7L, CIBIL: 650 → Medium Risk

Now classify:

Income: ₹9L, CIBIL: 720

BFSI use:

• Fraud / AML tagging

• Loan rejection reason classification

• Customer intent detection

⚠️ Trade-off: More tokens → higher cost & latency

4️⃣ Role-Based Prompt

What it is:You tell the model who it is

When to use:

• Expert-level responses

• Regulated domains

• Tone & depth control

Example:

“You are a senior credit risk officer in an Indian bank…”

Enterprise use:

• Credit policy explanations

• Compliance responses

• Executive summaries

🎯 Very common in production

5️⃣ Structured / Format-Constrained Prompt

What it is:Strict output format (JSON, table, schema)

When to use:

• API integration

• Downstream automation

• Data pipelines

Example:

“Return output strictly in JSON with fields: risk_score, reason, confidence”

Enterprise use:

• Decision engines

• Workflow orchestration

• RPA + GenAI

✅ Mandatory for production systems

6️⃣ Chain-of-Thought Prompt (Guided Reasoning)

What it is:Asks model to reason step-by-step

When to use:

• Complex decisions

• Multi-criteria evaluation

• Debugging logic

Example:

“Evaluate loan eligibility step by step considering income, credit score, liabilities.”

BFSI use:

• Credit decisions

• Risk scoring explanations

⚠️ In production:Often replaced with hidden reasoning + final answer to avoid hallucination or data leakage.

7️⃣ Retrieval-Augmented Prompt (RAG)

What it is:Prompt + external knowledge (documents, DB, policies)

When to use:

• Enterprise knowledge

• Compliance-safe answers

• Avoid hallucinations

Example:

“Answer using ONLY the provided RBI circular text.”

Enterprise use:

• Policy Q&A

• Internal SOP bots

• Regulatory assistants

🔥 Most important for BFSI GenAI systems

8️⃣ Instruction-Refinement / Self-Correction Prompt

What it is:Model reviews or improves its own output

When to use:

• Quality control

• Critical outputs

• Client-facing content

Example:

“Review the above response for regulatory risk and correct it.”

Enterprise use:

• Legal review

• Proposal drafting

• Customer communication

9️⃣ Tool-Calling / Function-Calling Prompt

What it is:Model decides which function or API to call

When to use:

• Orchestration

• Automation

• Agentic workflows

Example:

“If KYC is incomplete, call verify_kyc(). If complete, call credit_score().”

Enterprise use:

• Loan workflows

• Customer service automation

• Decision engines

🔟 Guardrail / Safety Prompt

What it is:Explicit constraints and boundaries

When to use:

• Regulated industries

• Customer-facing AI

• Risk mitigation

Example:

“Do not provide financial advice. Escalate to human if confidence < 0.7.”

BFSI use:

• Customer chatbots

• Advisory systems

• GenAI governance

📊 Summary Table (Interview-Ready)

Prompt Type	Used For	BFSI Example
Zero-Shot	Simple tasks	Email summary
Few-Shot	Classification	Fraud tagging
Role-Based	Expertise	Credit officer bot
Structured	Automation	Risk JSON output
Chain-of-Thought	Reasoning	Loan eligibility
RAG	Accuracy	RBI policy Q&A
Tool-Calling	Orchestration	Loan workflow
Guardrail	Safety	Compliance chatbot

🎯 How to Explain as a VP / Architect

“In enterprise GenAI, we don’t rely on a single prompt type.We layer prompts — role-based + RAG + structured output + guardrails — to ensure accuracy, compliance, and automation readiness.”