문의를 보내주셔서 감사합니다! 팀원이 곧 연락드리겠습니다.
예약을 보내주셔서 감사합니다! 저희 팀 멤버 중 한 분이 곧 연락드리겠습니다.
기간 7 시간
코스 개요
- Day 1: From a document collection to a knowledge agent
- How AI answers from documents
- Searching the collection and answering with a source
- Why answers are missing or wrong
- Difference between uploaded files and connected file storage
- Exercise: Collect three typical questions from the area that the agent should answer.
- Preparing knowledge
- Structure, naming and currency
- Removing duplicates and contradictory versions
- Checking permissions
- Exercise: Review the own document collection and put together a cleaned selection.
- Choosing a platform
- Copilot agent on SharePoint
- Claude project or custom GPT
- Langdock assistant and NotebookLM
- Criteria such as data storage, sharing, size of the collection and cost
- Exercise: Choose the right platform for the own case and justify the choice.
- Building the agent
- Instructions with audience, tone and answer format
- A source reference in every answer
- Handling questions without an answer in the collection
- Exercise: Build the knowledge agent on the chosen platform.
- Testing with a question catalog
- 20 test questions with expected answers
- Rating answers as correct, partly correct, wrong and missing
- Improving instructions and collection in a targeted way
- Exercise: Test the agent with 20 questions and record the hit rate.
- Maintenance and operation
- Responsibilities for content
- Updates and version status
- Collecting feedback from users
- Exercise: Write a maintenance plan with responsibilities and rhythm.
- Outcome: Each participant has a shared knowledge agent on their own documents, a measured hit rate and a maintenance plan.
- How AI answers from documents
요건
- Access to Copilot, Claude, ChatGPT, Langdock or NotebookLM, and a collection of released documents from the own area.
- No programming skills required.
Audience
- Teams, knowledge owners and key users who want to provide answers from their own documents, policies and manuals.
