SuperAPI AI Directory
Partner中文AI newsFavorites
Skillstriple-due-diligence

triple-due-diligence

Office & productivity

View repository

Updated 10-16 10:22

About this skill

Source Metadata

Original cangjie-skill frontmatter from the distillation run:

name: triple-due-diligence
description: |
  当用户准备为AI项目投入大量资源，需要全面评估技术可行性、商业价值和伦理影响时使用。
  用户在问"这个项目值得做吗"、"该投入多少"、"技术上能做吗"、"有没有风险"时激活。
  不适用于: 小型探索性项目（几天工作量）、纯技术研究、已明确无商业或伦理维度的场景。
source_book: 《给所有人的AI入门课》 吴恩达 (Andrew Ng)
source_chapter: p15, p06, p24
tags: [due-diligence, project-selection, feasibility, ethics, risk-assessment]
related_skills:
  - slug: ml-feasibility
    relation: depends-on
  - slug: three-value-drivers
    relation: depends-on
  - slug: cross-functional-brainstorming
    relation: composes-with
Triple Due Diligence — Technical × Business × Ethical

R — Original Text (Reading)

"Technical due diligence is to ensure that the AI system you hope to build is truly achievable... Business due diligence ensures that the envisioned project is valuable to the business... I also want to conduct ethical due diligence."

— Andrew Ng, p15

I — Methodological Skeleton (Interpretation)

Before investing months of resources in an AI project, three independent due diligences should be conducted in parallel: First, technical due diligence — consult AI technical experts to confirm whether the system can achieve the expected performance metrics? Is there enough data? How long is the development cycle? Where are the technical risks? Second, business due diligence — build a financial model to quantify how much value the project can create? Is it reducing costs, increasing revenue, or opening up new business? What is the input-output ratio? Third, ethical due diligence — even if both technical and business pass, still ask: Is this project truly beneficial to society? Will it harm specific groups? Does it align with long-term values? All three investigations must pass for the project to get a "green light". If any one fails, it should be reconsidered or abandoned. The core value of this framework is: before enthusiastic investment, use a structured approach to avoid the trap of "technically feasible, commercially should do, but ethically should not do".

A1 — Application in the Book (Past Application)

Case 1: Speech Recognition System

  • Problem: Whether it is worth developing a speech recognition system
  • Use of the Methodology:
    • Technical: Can it achieve 95% recognition accuracy? Is there enough speech data?
    • Business: How much manual transcription cost can it save? What new application scenarios can it open up?
    • Ethical: Is recognition fair across different accents? Will it exacerbate the digital divide?
  • Conclusion: All three need independent evaluation; if any one fails, it should be reconsidered
  • Result: Became a standard evaluation template for AI product initiation

Case 2: Visual Quality Inspection System

  • Problem: Whether to deploy AI visual quality inspection on the production line
  • Use of the Methodology:
    • Technical: Can it achieve 99% detection accuracy? Is the missed detection rate acceptable?
    • Business: How many quality inspectors can it replace? How long is the investment payback period?
    • Ethical: How should displaced workers be resettled? Are the detection standards biased against certain product types?
  • Conclusion: Even if it can technically achieve 99%, if the ethical impact on workers is too great, it should be advanced cautiously
  • Result: Emphasizes that AI deployment needs to balance efficiency and fairness

A2 — Trigger Scenarios (Future Trigger) ★

In what situations will users need this skill?

  1. An enterprise is preparing to launch an AI project that requires months of time and substantial resources
  2. Decision-makers are reviewing an AI project proposal and need a comprehensive evaluation framework
  3. Investors are evaluating the project feasibility of an AI startup
  4. The team has completed preliminary technical validation and is ready to enter the formal development stage.
  5. The project involves sensitive data or may affect the rights and interests of specific groups.

Language signals (activate when these appear in the user's words)

  • "Is this project worth doing?"
  • "How much should be invested?"
  • "Is it technically feasible?"
  • "Are there any risks?"
  • "What is the ROI?"
  • "Will there be any negative impact?"

Distinction from adjacent skills

  • With ml-feasibility difference: the two-factor framework is a rapid technical feasibility screening (concept + data), while triple due diligence is an in-depth comprehensive assessment (technical + commercial + ethical), suitable for formal project initiation decisions.
  • With cross-functional-brainstorming difference: cross-functional brainstorming is creative discovery during the project selection stage, while triple due diligence is in-depth assessment after the project is determined.
  • With automate-task difference: task decomposition is determining the project scope, while triple due diligence is assessing whether the project is worth investing in.

E — Executable Steps (Execution)

After the skill is activated, the agent should execute the following steps:

  1. Technical Due Diligence

    • Advise the user to consult an AI technical expert and answer the following questions:
      • What performance level can the system achieve? (accuracy, recall, latency, etc.)
      • Is there enough high-quality annotated data?
      • What are the development cycle and key technical risks?
      • What technical infrastructure and team capabilities are needed?
    • Completion criteria: The user obtains a clear technical feasibility conclusion and a risk list.
    • Stopping condition: If it is clearly technically infeasible (e.g., the required accuracy far exceeds the current SOTA), directly recommend terminating the project.

  2. Commercial Due Diligence

    • Advise the user to build a financial model and answer the following questions:
      • How much value can the project create? (cost savings, revenue growth, new business opportunities)
      • What are the input costs? (development, data, deployment, maintenance)
      • How long is the investment return period?
      • What is the opportunity cost of not doing it?
    • Completion criteria: The user obtains a quantified input-output analysis and investment return expectation.
    • Stopping condition: If the business model shows a negative ROI or an excessively long payback period, recommend redesign or termination.

  3. Ethical Due Diligence

    • Advise the user to further ask the following questions:
      • Is this project fair to all affected groups?
      • Could it harm specific vulnerable groups?
      • If this project is negatively reported by the media, how would you respond?
      • Does it align with your personal/organizational long-term values?
    • Completion criteria: The user has a clear understanding of the project's social impact and can give a "beneficial/harmful/needs modification" judgment.
    • Stopping condition: If the ethical assessment finds serious negative impact that cannot be mitigated, the project should be abandoned even if both technical and commercial assessments pass.

  4. Comprehensive Decision

    • All three pass → recommend launching the project.
    • Any one fails → recommend modifying the failed item, or terminating the project.
    • Completion criteria: The user obtains a clear "proceed/modify/terminate" decision recommendation.

B — Boundary (Boundary) ★

Do not use this skill in the following situations

  • Very small projects (a few days of work) — just try them quickly; formal due diligence is not needed
  • Purely technical research (with no clear commercial application or social impact)
  • Mature application scenarios where technical feasibility is already fully established
  • Rapid decision-making in emergencies (triple due diligence takes time)

Failure modes the author warns about in the book

  • Doing only technical and commercial due diligence while skipping ethical assessment — "it can make money and is technically feasible" does not equal "it should be done"
  • Ethical due diligence becomes a formality — no real probing of sharp questions, just going through the motions
  • The three investigations are completed by the same person — lacking an independent perspective and prone to confirmation bias

The author's blind spots / limitations of the era

  • Ethical due diligence is most often skipped in practice — the framework emphasizes its importance but does not provide detailed operational methods and evaluation criteria
  • The ideal execution order and weighting of the three due diligences are not clear — in reality, technical is often done first, commercial second, and ethical last (or not at all)
  • The book does not discuss "the cost of due diligence" — for small companies, comprehensive due diligence itself may be a heavy burden
  • Ethical standards are culturally relative — the book is mainly from a Western perspective, and ethical judgments may differ across cultural backgrounds

Adjacent methodologies that are easily confused

  • Difference from "investment due diligence": investment due diligence focuses on financial and legal risks, while triple due diligence adds technical and ethical dimensions
  • Difference from "Impact Assessment": impact assessment usually focuses only on negative social impact, while triple due diligence also considers commercial value

Related skills

  • ml-feasibility (depends-on): The technical due diligence in triple due diligence is built on the basis of passing the two-factor framework assessment
  • three-value-drivers (depends-on): Commercial due diligence requires quantifying project value, and the three-value driver framework provides a quantification method
  • cross-functional-brainstorming (composes-with): After brainstorming produces candidate projects, use triple due diligence for in-depth pre-project evaluation

Audit information

  • Verification passed: V1 ✓ / V2 ✓ / V3 ✓
  • Test pass rate: Pending testing
  • Distillation time: 2026-06-29