# How Management-Deep-Dive Evaluates Corporate Leadership Using the 'Buy the Person' Principle

> Discover how the management-deep-dive skill in AI-Berkshire evaluates corporate leadership. Learn the 9-step workflow for scoring Integrity, Ability, and Trustworthiness using the 'buy the person' principle.

- Repository: [Xbt Lin/ai-berkshire](https://github.com/xbtlin/ai-berkshire)
- Tags: deep-dive
- Published: 2026-07-28

---

**The `/management-deep-dive` skill in AI-Berkshire transforms the "buy the stock = buy the person" investment mantra into a data-driven, 9-step workflow that scores leadership across three pillars: Integrity, Ability, and Trustworthiness.**

The `management-deep-dive` skill in the [xbtlin/ai-berkshire](https://github.com/xbtlin/ai-berkshire) repository implements a structured framework for evaluating corporate leadership based on the classic value investing principle that purchasing a stock means entrusting your capital to the person running the business. Drawing from the philosophies of 段永平 (Duan Yongping), 巴菲特 (Warren Buffett), and 李录 (Li Lu), this skill quantifies qualitative leadership traits into reproducible scores that feed directly into the AI-Berkshire investment decision pipeline.

## The "Buy the Person" Philosophy in AI-Berkshire

The canonical skill definition in [[`skills/management-deep-dive.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/management-deep-dive.md)](skills/management-deep-dive.md) organizes leadership evaluation around three thematic pillars that map directly onto the "buy the person" principle. Each pillar addresses a fundamental question: Is the person honest? Does the person have ability? Would you entrust them with your money for a decade?

## The Three Pillars of Leadership Evaluation

### Integrity: The "First-Vote Killer"

**Integrity** serves as the foundational filter—management without integrity is disqualified immediately regardless of other capabilities.

The skill measures integrity through:

- **Commitment vs. Delivery Analysis**: A tracking table that monitors promises made during earnings calls, shareholder letters, and interviews against actual outcomes, calculating a **兑现率** (fulfillment rate). A score above 80% earns an "优秀" (excellent) rating.
- **Crisis Handling Assessment**: Evaluation of how leadership teams responded to major setbacks and unforeseen challenges.
- **Stakeholder Attitude Survey**: Aggregation of feedback from employees, customers, and suppliers regarding ethical behavior.

### Ability: Strategic and Execution Capacity

**Ability** encompasses the capacity to generate returns through strategic foresight and operational excellence.

The evaluation includes:

- **CEO Ability-Circle**: Analysis of strategic vision (accuracy of past forecasts) and execution capability (strategy-to-implementation gaps).
- **Forecast Accuracy Scoring**: Historical predictions are matched against outcomes to compute quantitative accuracy metrics.
- **Capital-Allocation Record**: Detailed scoring of M&A decisions, share repurchases, dividends, and new-business investments using a 1-5 scale with supporting rationale.

### Trustworthiness: The 10-Year Entrustment Test

**Trustworthiness** determines whether an investor could comfortably hand over capital for a decade without daily oversight.

Key measurements include:

- **Succession Scenario Analysis**: Assessment of whether the business would survive and thrive if the current CEO departed tomorrow.
- **Governance Assessment**: Review of equity structure risks, insider ownership levels, AB-share arrangements, and compensation alignment with long-term shareholder value.
- **Side-Validation**: Cross-referencing employee reviews (Glassdoor, Zhihu), customer app-store ratings, and industry forum sentiment to verify internal assessments.

## The 9-Step Evaluation Workflow

The skill executes a parallelized, multi-agent workflow defined in [[`skills/management-deep-dive.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/management-deep-dive.md)](skills/management-deep-dive.md):

1. **Identify Key Executives** – Populate a roles table (CEO, CFO, Founder, controlling shareholders) via **WebSearch**.
2. **Parallel Data Collection** – Four background agents concurrently scrape: (a) CEO public statements, (b) capital-allocation decisions, (c) governance data, and (d) side-validation sources.
3. **CEO Ability-Circle Evaluation** – Compute forecast accuracy and judge execution on "say → do" consistency, talent attraction, and crisis response.
4. **Integrity Assessment** – Tabulate promises with fulfillment status (✅/⚠️/❌) and calculate the **承诺兑现率** (commitment fulfillment rate).
5. **Capital-Allocation Scoring** – Generate a **资本配置评分** (capital allocation score) matrix for each major financial decision.
6. **Governance & Compensation Review** – Examine equity structures, voting rights, and insider ownership alignment.
7. **Stakeholder Attitude Validation** – Aggregate sentiment from public employee and customer feedback channels.
8. **Succession & "Buy-for-10-Years" Test** – Scenario analysis testing operational continuity without the current CEO.
9. **Report Generation** – Produce a structured markdown report at `reports/{company}-management-{YYYYMMDD}.md` containing a **综合评分** (composite score).

## Practical Usage and Output Format

Invoke the skill using simple natural language commands:

```bash

# Basic usage – company only

/management-deep-dive 美团

# Include named executive for targeted analysis

/management-deep-dive 王兴 美团

```

These commands trigger the parallel agent workflow and generate a comprehensive report. The output includes structured evaluation tables such as:

```markdown

## 诚信度评估

| # | 时间 | 承诺内容 | 承诺场合 | 兑现情况 | 评价 |

|---|------|----------|----------|----------|------|
| 1 | 2022-Q4 | "2025年实现X业务盈利" | 2022年报电话会 | ✅ | 兑现率>80%→优秀 |

## 综合评分

| 维度 | 权重 | 评分 | 加权 |
|------|:----:|:----:|:----:|
| 诚信度 | 35% | 4.5 | 1.58 |
| 战略与执行能力 | 25% | 4.0 | 1.00 |
| 资本配置能力 | 25% | 3.8 | 0.95 |
| 治理结构 | 15% | 3.5 | 0.53 |
| **综合** | 100% | **4.1** | |

```

The composite scoring weights reflect the priority of integrity in value investing: **Integrity 35%**, **Strategic/Execution Ability 25%**, **Capital Allocation 25%**, and **Governance Structure 15%**.

## Summary

- The `/management-deep-dive` skill operationalizes the "buy the person" principle through a structured, file-based workflow defined in [`skills/management-deep-dive.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/management-deep-dive.md).
- Leadership evaluation rests on three pillars: **Integrity** (35% weight), **Ability** (25%), and **Trustworthiness** (embedded in Governance at 15% alongside Capital Allocation at 25%).
- The workflow employs parallel agents to collect data on public statements, capital allocation history, governance structures, and stakeholder sentiment.
- Quantified outputs include **兑现率** (fulfillment rates), **资本配置评分** (capital allocation scores), and a weighted **综合评分** (composite score) that feeds into the broader AI-Berkshire investment pipeline.
- Reports are generated as timestamped markdown files in the `reports/` directory, providing reproducible documentation for investment decisions.

## Frequently Asked Questions

### How does the skill calculate the integrity fulfillment rate?

The skill scans earnings calls, shareholder letters, and public interviews to extract specific commitments made by leadership. These promises are tracked in a commitment table with timestamps and contexts. As results become available, each item is marked as fulfilled (✅), partial (⚠️), or failed (❌). The **兑现率** is calculated as the percentage of fulfilled commitments; rates exceeding 80% receive an "优秀" rating according to the scoring logic in [`skills/management-deep-dive.md`](https://github.com/xbtlin/ai-berkshire/blob/main/skills/management-deep-dive.md).

### What data sources does the parallel collection step use?

Four background agents concurrently gather: (1) CEO public statements from earnings transcripts and interviews; (2) capital-allocation decisions from SEC filings and press releases; (3) governance and compensation data from proxy statements and ownership disclosures; and (4) side-validation from employee review platforms (Glassdoor, Zhihu), customer app-store ratings, and industry forum discussions. This multi-source approach reduces reliance on single-channel bias.

### Can the skill evaluate private companies or executives without public statements?

The skill relies heavily on publicly available data—earnings calls, regulatory filings, and media coverage. For private companies or low-profile executives, the WebSearch step may return limited results, resulting in incomplete **Ability-Circle** or **Integrity** tables. The framework will flag data gaps in the generated report rather than extrapolate scores, ensuring that evaluations remain grounded in verifiable evidence.

### How does the "Buy-for-10-Years" test work in practice?

The succession scenario analysis examines organizational depth, documented succession plans, and the concentration of operational knowledge in the current CEO. The skill asks whether the business model, culture, and operational systems are robust enough to survive the CEO's departure without significant value destruction. This includes evaluating whether a clear successor has been groomed and whether key executives have retention incentives aligned with long-term stability.