When you’re working in an Omnifact Space with uploaded documents, you’re using one of AI’s most powerful capabilities: getting information from your own documents. This means AI can search through your documents, find relevant information, and use it to provide accurate, grounded responses. Think of it like having a research assistant who has read all your company documents and can instantly find and cite the exact information you need for any question.

​
Understanding Document Search in Omnifact Spaces

​
What Happens Behind the Scenes

When you ask a question in a Space with uploaded documents:
  1. Your question is analyzed - AI understands what you’re looking for
  2. Documents are searched - The system finds relevant sections from your uploaded files
  3. Information is found - Key passages are pulled from the most relevant documents
  4. Response is created - AI creates an answer based on the found information
  5. Sources are provided - You see which documents were used

​
Why This is Powerful

  • Accuracy: Answers are based on your actual documents, not AI’s general knowledge
  • Current Information: Information is as up-to-date as your uploaded documents
  • Specific to You: Responses are tailored to your organization’s context
  • Verifiable: You can check answers by looking at the source documents

​
The Art of Knowledge Base Prompting

​
Basic Principles

​
1. Be Specific About Your Information Needs

Instead of asking broad questions, focus on specific information you need from your documents. ❌ Vague: “What’s our policy on vacation?”
✅ Specific: “What is the maximum number of consecutive vacation days an employee can take according to our HR policy?” ❌ Vague: “Tell me about the project”
✅ Specific: “What are the key milestones and deadlines outlined in the Project Alpha documentation?“

​
2. Use Document-Aware Language

Ask as if the answer is in your documents: ✅ “According to our quarterly report, which product lines grew most?”
✅ “From customer feedback, what are the top three feature requests?“

​
3. Request Evidence and Citations

Ask for sources in the answer: ✅ “What are our data retention requirements, and which policy document states them?”
✅ “List approved software vendors with the policy section for approval criteria.”

​
Structured Information Extraction

​
Creating Summaries from Multiple Sources

Create a comprehensive executive summary of our Q3 performance by combining information from:
- The quarterly financial report
- Customer satisfaction survey results  
- Employee engagement data
- Market analysis documentation

Structure the summary with: Financial Performance, Customer Insights, Employee Satisfaction, Market Position, and Key Recommendations.

​
Building Comparison Tables

Create a comparison table of our three main product lines using information from our product specification documents. Include features, target markets, pricing models, and competitive advantages for each product.

​
Extracting Action Items and Decisions

Review all meeting minutes from the leadership team for October and extract:
- All action items with assigned owners and deadlines
- Strategic decisions made and their rationale
- Unresolved issues that need follow-up

Present this as a consolidated action tracker.

​
Basic Troubleshooting

​
When AI Can’t Find Information

​
Problem: “I don’t see that information in the available documents”

Solutions:
  1. Rephrase your question using different terminology
  2. Break down complex questions into simpler parts
  3. Check if the information actually exists in your uploaded documents
  4. Try broader search terms first, then narrow down

​
Example Troubleshooting Sequence:

// First attempt - too specific
"What is our customer churn rate for the enterprise segment in Q3?"

// If not found, try broader
"What customer retention metrics are reported in our Q3 documents?"

// Then narrow down based on what's available
"Based on the retention data you found, can you calculate or identify metrics specifically for enterprise customers?"

​
When Retrieved Information Seems Incorrect

​
Verify and Cross-Reference

You mentioned that our customer satisfaction score is 85%. Can you show me exactly where this number appears in our documents and confirm it's referring to overall satisfaction rather than a specific metric?

​
Request Multiple Sources

Find all references to our employee turnover rates in our HR documents. If there are different numbers, explain the differences and which source is most recent.

​
When Responses Are Too General

​
Ask for Specific Details

You mentioned we have a "comprehensive training program." Can you provide the specific details from our HR documentation about training requirements, duration, and certification processes?

​
Request Step-by-Step Information

Instead of summarizing our incident response process, please provide the exact step-by-step procedure as outlined in our IT security documentation.

​
Best Practices for Knowledge Base Prompting

​
1. Start Broad, Then Narrow

Begin with general questions to understand what information is available, then drill down: 
// First: Explore what's available
"What topics are covered in our customer service documentation?"

// Then: Get specific information
"Based on that overview, what are the specific response time requirements for different types of customer inquiries?"

​
2. Use Follow-Up Questions Effectively

Build on previous responses to dig deeper: 
// Initial question
"What are our main revenue streams according to the financial reports?"

// Follow-up based on response  
"For the subscription revenue you mentioned, what are the growth trends and customer acquisition costs over the past year?"

​
3. Request Structured Output for Complex Information

Create a structured breakdown of our product development process from the engineering documentation:

**Phase 1: Planning**
- Key activities:
- Required approvals:
- Typical duration:

**Phase 2: Development**
- Key activities:
- Quality checkpoints:
- Typical duration:

[Continue for all phases]

​
Getting Started

Working with Knowledge Base documents transforms your organization’s information into an intelligent, searchable resource. Start with simple, specific questions about your uploaded documents, then build complexity as you get comfortable. Remember: precise prompts lead to precise answers.