pre-visit-plan

Pre-Visit Planning

Target Role

Role:Corporate Relationship Manager
Use Case:The preparation stage before visiting a customer, usually carried out 1-3 days before the visit
Output Purpose:To understand the full picture of the customer, set visit objectives, and prepare marketing talking points and product recommendation plans
Decision Level:Information assistance and strategic guidance, not involving credit approval decisions
Execution Frequency:Executed once before each customer visit

Data Sources

Required Data

Data Item	Source	Acquisition Method	Sensitivity Level
Customer basic information	In-house customer information system	API: /api/customer/basic	Internal
Deposit and AUM data	Core banking system	API: /api/customer/deposits	Internal
Credit utilization	Credit management system	API: /api/customer/credit	Confidential
Transaction records	Core banking system	API: /api/customer/transactions	Confidential
Historical due diligence/visit records	Image archive system	File reading	Internal
Risk alert information	Risk alert system/external business and judicial platforms	API/crawler	Public/Internal
Industry trends	Industry research knowledge base/external information	Retrieval/web search	Public

Data Masking Rules

Personal ID number: show the first 3 and last 4, replace the middle with *
Bank account number: show only the last 4 digits
Contact information: do not appear in the output
Customer sensitive financial data: use only in internal reports, must not be disclosed externally

Fallback Strategy

If risk alert data is unavailable: mark "risk alert dimension not included", and continue with the remaining analysis
If historical visit records are missing: mark "no historical visit records", and generate a first-visit strategy based on public information
If the industry data source times out: use public information for a brief analysis, and clearly mark "industry analysis based on public information"
If the customer's financial data covers only 1 year: mark "insufficient data, trend analysis unavailable", and perform only static analysis
If the customer 360 system is unavailable: mark "customer basic data not verified", and continue based on the relationship manager's verbal information
If the image archive system is unavailable: mark "historical documents not reviewed", and prompt the relationship manager to manually provide key information

Workflow

Step 0: Data Confirmation and Verification

List input parameters: customer name, visit date, visit purpose, special points of concern. Confirm the data time range (last 3/6/12 months) and the validity of the industry classification code. Run scripts/validate_visit_plan.py to check the completeness of input parameters.

✅ Validation passed → proceed to Step 1
❌ Validation failed → return the missing items list and ask the user to supplement

📋 Data source:user_upload(customer information provided by the user)

Step 1: Obtain core customer data

Query the customer 360-degree panoramic view to obtain:

Deposit scale and structure (proportion of demand deposits/time deposits/wealth management products)
Total credit line and utilization (used credit line, remaining credit line, guarantee method)
Recent transaction flows (transaction flow change trends over the last 3/6/12 months, seasonal characteristics, abnormal fluctuations)

Do not skip the data acquisition step; all numbers must show the calculation process.

📋 Data source:system_api(in-bank customer information system) 📋 Executing entity:ai 📋 Confirmation mechanism:none

Step 2: Enumerate customer profile files

List all available documents for this customer in the imaging system and file storage:

Due diligence reports (most recent 1-2 times)
Visit records (most recent 3 times)
Credit approval summary
Financial statements
Other related documents (contracts, invoices, etc.)

📋 Data source:system_api(imaging archive system) 📋 Executing entity:ai 📋 Confirmation mechanism:none

Step 3: Read key historical documents

Prioritize reading the contents of the following documents:

The most recent due diligence report (focus on risk points and credit recommendations)
The last visit record (focus on commitment follow-up status and customer feedback)
The last credit approval summary (focus on approval conditions and additional requirements)
If the user specifies particular documents, prioritize reading those documents

Do not skip the historical document reading step; key information must be extracted for subsequent analysis.

📋 Data source:system_api(file content reading interface) 📋 Executing entity:ai 📋 Confirmation mechanism:none

Step 4: Obtain risk warning signals

Query this customer's risk warning records within the last 90 days:

Judicial risk: person subject to enforcement, dishonest person subject to enforcement, equity freeze, consumption restriction order
Business risk: list of business abnormalities, administrative penalties, tax arrears announcements
Public opinion risk: negative news reports, impact of changes in industry policies

If any risk signal is found, it must be prominently marked at the beginning of the subsequent report, and it must not be ignored by continuing.

📋 Data source:system_api(risk warning system) or external business/judicial data platforms 📋 Executing entity:ai 📋 Confirmation mechanism:none

Step 5: Obtain industry background information

Retrieve recent developments in the industry to which this customer belongs:

Industry policy changes (within the last 3 months)
Market prosperity index
Changes in the competitive landscape
Fluctuations in raw material prices (if applicable)
Focus on information from the last 3 months

Consult references/industry-classification.md to obtain precise industry classification characteristics.

📋 Data source:reference(in-bank industry knowledge base) or online search 📋 Executing entity:ai 📋 Confirmation mechanism:none

Step 6: Comprehensively assess the visit request

Integrate the above data, analyze from 4 dimensions, and identify the core type of this visit:

Customer Fundamentals Dimension:

Enterprise size, industry classification (GB/T 4754-2017 refined to the sub-class level), years of operation, equity structure
Background and credit status of major shareholders/actual controllers
Relationship with the Bank: first loan time, years of cooperation, determination of lead bank/non-lead bank
Deposit change trends, credit expiration time

Recent Developments Dimension:

Transaction flow change trends, abnormal fluctuations
Risk warning signals
Changes in core financial indicators (revenue, profit, debt ratio)
Impact of industry policy changes on market prosperity

Marketing Opportunity Dimension:

Credit expiration and loan renewal timeline
New financing needs (investment expansion/production expansion/M&A/working capital supplementation)
Deposit growth potential (temporary fund deposits/payroll agency/supply chain funds)
Intermediary business opportunities (settlement account opening/bills/foreign exchange/wealth management/asset management)
Opportunities for joint development with upstream and downstream supply chain

Relationship Status Dimension:

In-depth assessment of relationships with decision-makers (Chairman/CFO/Finance Director)
Business distribution with competing banks
Historical visit records and commitment follow-up status
Customer satisfaction and potential dissatisfaction points

Visit Type Determination:

First visit
Regular maintenance
Loan renewal follow-up
Risk investigation
Marketing advancement

Step 7: Personalized Strategy Generation

Based on the visit type and customer characteristics, select a differentiated strategy template:

Refer to references/visit-strategy-guide.md to obtain the visit strategy guide.

Visit Type	Focus	Key Data Focus
Loan renewal follow-up	Credit expiration milestone, funding gap analysis, collateral status	Expiration time, transaction flow health, interest rate market conditions
Risk investigation	Verification of risk signals, actual operating conditions, repayment sources	Judicial records, transaction flow changes, accounts receivable collection
New marketing	Needs discovery, competing bank distribution, decision-making chain	Recent fund usage, capital expenditure plans
Relationship maintenance	Satisfaction, potential dissatisfaction, cross-selling opportunities	AUM changes, intermediary business volume changes
First visit	Basic information collection, relationship building, needs assessment	Industry background, publicly available enterprise information

Adjust the output content according to the customer's specific circumstances, and generate a complete pre-visit planning report.

📋 Executing entity:ai 📋 Confirmation mechanism:inform(Notify the account manager for confirmation after generation)

Step 8: Report Formatting Output

Use the assets/visit-plan-template.md template for formatted output. Ensure all sections are complete, and data is annotated with sources and dates. Risk warnings (if any) must be placed in a prominent position at the beginning of the report.

✅ Format complete → Proceed to Step 9
❌ Format missing → Supplement the missing sections and output again

Step 9: Final Validation and Delivery

Run scripts/validate_visit_plan.py Validate output report compliance:

Check completeness of required sections
Verify number of primary objectives (1-3)
Check industry classification precision (mid-level or sub-level)
Verify risk front-loading (if there are risk signals)
Check data timeliness annotation
✅ Validation passed → output final report
❌ Validation failed → fix issues and re-validate

📋 Executing entity:ai 📋 Confirmation mechanism:confirm(requires customer manager confirmation before delivery)

Output Format

Use the assets/visit-plan-template.md template. The report must include the following sections:

# 访前规划报告

**客户名称**:[企业全称]
**拜访日期**:[建议日期或用户指定日期]
**规划生成时间**:[当前日期]
**本次拜访类型**:[首次拜访 / 定期维护 / 续贷跟进 / 风险排查 / 营销推进]

---

## ⚠️ 风险提示(如无风险信号则省略此节)

> [醒目标注近期风险信号,如:XX公司已被列入失信被执行人名单(日期),建议拜访前与风险部门确认处置预案]

---

## 一、客户基本面回顾

### 基础信息
| 项目 | 内容 |
|------|------|
| 行业分类 | [GB/T 4754-2017 细化至小类] |
| 企业规模 | [大型/中型/小型/微型] |
| 实控人 | [姓名 + 简述背景] |
| 与本行关系 | [合作年限 / 主办行状态] |

### 授信与存款现状
| 项目 | 金额/情况 | 备注 |
|------|-----------|------|
| 授信总额 | | |
| 已用额度 | | |
| 最近到期时间 | | |
| AUM规模 | | 近6个月变化趋势 |

---

## 二、近期动态分析

### 经营动态
[基于流水变化和财务数据的客观描述,包含具体数字和时间节点]

### 行业背景
[该行业近期1-2个关键动态,说明对该企业的潜在影响]

### 风险信号扫描
[列出所有司法、舆情、经营异常信号;如无则标注"未发现异常信号(截止日期)"]

---

## 三、本次拜访目标

**主目标**(必须完成):
1. [具体、可量化的目标,例如:确认XX贷款续贷意向,了解资金缺口规模]

**次目标**(争取完成):
2. [次要目标]
3. [次要目标(可选)]

**信息收集任务**:
- [需要了解的关键信息,例如:新项目投资进度、竞争行报价水平]

---

## 四、材料准备清单

### 需携带材料
- [ ] [具体材料名称,例如:最新授信产品方案(含利率区间)]
- [ ] [...]

### 需提前确认
- [ ] [内部状态确认事项,例如:XX存量授信的风险部意见]
- [ ] [...]

### 需预约/准备
- [ ] [拜访前需完成的准备事项]

---

## 五、沟通策略

### 开场建议
[根据关系深度和拜访目的,给出具体的开场话术建议,1-3句]

### 核心沟通要点
1. **[话题一]**:[具体沟通策略和预期引导方向]
2. **[话题二]**:[...]
3. **[话题三]**:[...]

### 敏感话题应对
| 可能触及的敏感点 | 建议应对策略 |
|-----------------|-------------|
| [如:竞争行的更低利率] | [如:强调综合服务价值和关系稳定性] |

### 预期成果与下一步
- **本次拜访期望产出**:[例如:获得续贷授权/获得财务报告/确认追加担保意向]
- **建议下一步行动**:[例如:一周内发送授信方案 / 安排总行专家团队上门交流]
- **建议跟进时间节点**:[...]

All data annotations: data source + data date + whether it is the latest data.

This output can be parsed and used by visit-memo and submit-credit-application. Key fields include:

Customer name (string)
Visit type (enum: first visit/regular maintenance/loan renewal follow-up/risk investigation/marketing advancement)
Primary objectives (array[string], 1-3)
Risk alerts (array[string], optional)

The output must include a disclaimer, citing the shared/disclaimer-template.md template.

Compliance Constraints

Prohibition of return commitments: Under no circumstances may commitment-style expressions such as "expected to obtain X ten thousand in deposits" or "guaranteed successful marketing" be given.
Prohibition of data guessing: Missing data = ask the user for it or mark it in the report as "data missing". It is strictly prohibited to substitute industry averages for actual data.
Data timeliness: If the data is more than 6 months old, "⚠️ Data may be outdated" must be prominently marked at the beginning of the report.
Risk front-loading: If risk signals are found, they must be prominently marked at the beginning of the report and must not be buried in the main text.
Clear objectives: Each visit must set 1-3 primary objectives and must not list them in a vague, general way.
Personalization: The report content must reflect the specific characteristics of that customer. Use of generic template wording is prohibited.
Prohibition of unauthorized recommendations: This Skill only provides visit strategy recommendations and does not involve credit approval decisions or limit approval.
Prohibition of skipping steps: No step in the Workflow may be skipped, and they must be executed in order.
Mandatory red-line enforcement: If any red line (R1-R3) is triggered, it must be prominently marked at the beginning of the report and subsequent analysis must be paused.

Financial Compliance Red Lines

R1: Prohibition of false marketing - Do not exaggerate product returns, conceal product risks, or fabricate nonexistent marketing policies. Violating this will lead to regulatory penalties.

R2: Prohibition of leaking customer privacy - The report must not contain sensitive personal information such as unmasked ID numbers, complete bank account numbers, or contact information. Violating this will violate the Personal Information Protection Law.

R3: Prohibition of concealing major risks - If it is found that the customer has been listed as a judgment debtor for dishonesty, is involved in major litigation, or has abnormal operations, this must be prominently marked at the beginning of the report and the risk must not be downplayed for marketing purposes. Violating this will violate the principle of prudent operation.

Audit Trail

After each pre-visit planning session ends, generate an audit log audit/{客户简称}_{日期}_audit.json:

{
  "skill_name": "pre-visit-plan",
  "skill_version": "2.0.0",
  "execution_time": "2026-05-05T10:30:00+08:00",
  "input_params": {
    "customer_name": "XX企业",
    "visit_date": "2026-05-08",
    "visit_purpose": "续贷跟进"
  },
  "model": "claude-opus-4-7",
  "operator": "张三(工号:10086)",
  "steps": [
    {
      "step": "数据确认与验证",
      "executor": "ai",
      "data_source": {"type": "user_upload"},
      "result": "pass"
    },
    {
      "step": "获取客户核心数据",
      "executor": "ai",
      "data_source": {"type": "system_api", "system": "客户360系统"},
      "result": "pass"
    },
    {
      "step": "获取风险预警信号",
      "executor": "ai",
      "data_source": {"type": "system_api", "system": "风险预警系统"},
      "result": "pass"
    },
    {
      "step": "生成访前规划报告",
      "executor": "ai",
      "data_source": {"type": "context"},
      "confirmation": {"type": "confirm", "confirmed_by": "张三(工号:10086)", "confirmed_at": "2026-05-05T10:35:00+08:00"},
      "result": "pass"
    }
  ],
  "warnings": ["部分财务数据超过3个月"],
  "references_used": ["references/visit-strategy-guide.md", "references/industry-classification.md"]
}

Audit log retention period: at least 3 years.

Gotchas

#1: Industry classification too broad

Symptom: The industry background analysis only says "manufacturing" or "wholesale and retail", lacking specificity
Cause: The GB/T 4754-2017 classification was not refined to the medium or small category
Solution: Always query the customer's precise industry code and refer to references/industry-classification.md to obtain the sub-industry characteristics

#2: Historical visit records ignored

Symptom: The generated visit strategy is disconnected from the commitments made in the last visit, repeatedly asking for already-known information
Cause: Historical visit records and due diligence reports in the imaging system were not read first
Solution: Workflow Step 3 must be executed, reading the last 3 visit records and the most recent due diligence report

#3: Risk signals buried in the body text

Symptom: The customer has been listed as a dishonest judgment debtor, but the risk warning was not placed at the beginning of the report
Cause: The content was not prioritized when generating the report
Solution: After the risk signal scan is completed, if there is any anomaly, it must be marked in a separate section at the beginning of the report

#4: Visit objectives too vague

Symptom: Objectives are written as "understand the customer's situation" or "advance business cooperation", which cannot be verified as achieved
Cause: Objectives were not made specific and quantifiable
Solution: Each main objective must include specific actions and verifiable results, for example "confirm the intention to renew the XX loan and understand the scale of the funding gap"

#5: Data timeliness not annotated

Symptom: Financial data from 6 months ago was used, but the report did not note that the data may be outdated
Cause: Data timestamps and timeliness were not checked
Solution: During Step 0 data confirmation, the data time range must be verified, and expired data must be prominently marked at the beginning of the report

Examples

Example 1: Loan renewal follow-up visit

User input:

请为XX制造企业的拜访生成访前规划报告。客户授信5000万将于2026-06-15到期,需要跟进续贷事宜。拜访日期:2026-05-10。

Skill execution flow:

Step 0: Validate input parameters (customer name, visit date, visit purpose)
Step 1: Query customer 360 data → Obtain credit usage, deposit changes, transaction flow trends
Steps 2-3: Read historical files → Obtain the last due diligence report and approval summary
Step 4: Risk scan → No abnormal signals found
Step 5: Industry analysis → Recent favorable policies for manufacturing
Steps 6-7: Generate report → Visit type identified as "loan renewal follow-up", highlighting the loan renewal timeline and funding gap analysis
Steps 8-9: Format output and validate → Output the final report

Output key points:

Main objectives: Confirm loan renewal intention, understand the latest funding needs, collect the latest financial statements
Risk warning: None
Material preparation: Latest credit product plan, interest rate comparison analysis
Communication strategy: Start with the customer's business highlights and naturally transition to the loan renewal topic

Example 2: Risk investigation visit