0%
Inside SeriesConcept-Based Inquiry

An Assistant That Knows the Company Files

Internal Q&A is only as good as the passages it retrieves.

Who this book is for

Người làm công nghệ và AI trong doanh nghiệp Việt Nam: nhân viên IT một mình hay trong nhóm nhỏ, lập trình viên, người dựng tính năng AI, trưởng nhóm kỹ thuật ở công ty vừa và nhỏ; vừa phải giữ hệ thống chạy ổn, vừa được giao đưa công cụ và AI mới vào công việc của các phòng khác.

Sound familiar?

  • The assistant answers from retired procedures.
  • Documents are scattered across drives, email and unreadable scans.
  • Users phrase questions differently from the documents.
  • No one can check which passage an answer came from.

After reading, you can

Help AI builders build internal document Q&A that answers correctly and cites sources.

Each chapter answers one question

  1. Chapter 1: Is the assistant wrong, or are the documents stale?

    Which company documents are the current valid versions?

    An assistant only knows what is in the store, including what is outdated.

  2. Chapter 2: Can one bad cut ruin a whole answer?

    How long should a document chunk be to answer one question?

    How you chunk decides whether an answer is complete or cut off midway.

  3. Chapter 3: Bad model, or the wrong passage retrieved?

    When the assistant answers wrongly, is the fault in retrieval or in writing?

    Most wrong answers start from retrieving the wrong chunk, not from the model.

  4. Chapter 4: Can you trust a smooth answer with no source?

    Can users verify the assistant's answers?

    An answer that does not cite its source leaves readers no way to check.

  5. Chapter 5: If nobody owns the documents, who keeps them true?

    When a process changes, how soon does the assistant know?

    If the store is not updated, the assistant gives correct answers about a company that has changed.