Preparation for AI Chatbot Development

Building SmartWeb’s AI chatbot primarily requires “knowledge base preparation” and “establishing operational systems.”

Three Methods for Knowledge Base Construction

The information source (knowledge base) for AI chatbot responses can be built through the following three methods:

1. Website Crawling

Regularly crawl your company’s website to automatically gather the latest information.

  • Full domain crawling
  • Specific page designation
  • Sitemap-based crawling
  • Periodic automatic update settings

2. Q&A List Registration

Manually register and manage frequently asked questions and answers.

  • Utilize customer portal knowledge base
  • Immediate reflection
  • Prioritized presentation as responses

3. Document Upload

Upload existing materials directly.

Supported file formats:

  • PDF documents
  • Word documents (DOCX/DOC)
  • Excel (XLS/XLSX)
  • Text files (TXT)
  • CSV files
  • Markdown
  • HTML
  • YouTube video URLs

Materials to Prepare

Essential Materials

  • FAQ content: Frequently asked questions and answers (minimum 50 items recommended)
  • Product/service information: Basic explanatory materials
  • Knowledge base articles and internal manual documents
  • Product catalogs and specifications (PDF format recommended)
  • Company introduction materials and presentation documents
  • Video content (YouTube video URLs)
  • Existing support articles and documentation

Operational Preparation

Establishing Operational Systems

  • Designating operational staff: 1-2 people recommended
  • Setting response scope: Determining question categories the chatbot will handle
  • Escalation conditions: Designing handoff timing to human operators

Things to Decide

  • Policy for handling questions the AI cannot answer
  • Response policy for after-hours inquiries
  • Methods for monitoring response quality

Content Quality

AI chatbot response accuracy heavily depends on knowledge base quality.

Conditions for Good Content

  • Accurate and current information: Outdated information affects response accuracy
  • Clear writing: Avoid technical terms or provide explanations
  • Comprehensiveness: Cover common question patterns

Content Creation Tips

  • Keep sentences short (15-20 words as guideline)
  • Use headings and bullet points for structure
  • Avoid ambiguous expressions
  • Use active voice

Knowledge Base Creation Service

SmartWeb also provides support services for knowledge base creation.

  • Create question and answer drafts from past inquiries, homepages, manuals, product catalogs, etc.
  • Cost: 50,000 yen and up

Important notes:

  • Customer review and structuring of draft content required
  • Input into support portal is customer-side work

Implementation Process

  1. Requirements definition hearing: Clarify business goals and response scope
  2. Knowledge survey: Inventory existing documents and FAQs
  3. System construction: FlowHunt environment setup and LiveAgent integration configuration
  4. Knowledge base preparation: Crawl settings, document uploads, Q&A registration
  5. Testing and tuning: Verify and adjust response quality
  6. Launch operations: Gradual rollout and monitoring