Time required: Approximately 5 minutes reading time
Prerequisites:
Access to Captured Knowledge
Expected results: After reading this article, you will understand what the AI Knowledge Base does, how it affects the suggestions you see, and when to request updates to improve accuracy.
The AI keeps suggesting the wrong electrician for Building 3. You've corrected it five times this week. Next week, you'll probably correct it five more times.
This happens when the Knowledge Base doesn't know your preferences. The Knowledge Base is what makes the AI smart about your specific business - and understanding it helps you get better suggestions with less manual correction.
What is the Knowledge Base?
The Knowledge Base is your company's specific information that guides how the AI processes emails, creates cases, and suggests workflows. It contains details about:
How you handle different property management issues (plumbing, heating, electrical, etc.)
Your preferred service providers and when to use them
Your communication standards and procedures
Building-specific requirements and rules
How you want issues categorized and prioritized
Think of it as teaching the AI your company's playbook so it can make suggestions that match your established practices.
How the Knowledge Base Helps You
The Knowledge Base directly affects the quality of AI suggestions you receive every day.
Better Categorization
Without proper Knowledge Base content, the AI might categorize heating issues as routine year-round. With your knowledge, the AI understands that heating issues are urgent November through March in Austria, automatically prioritizing them correctly.
Accurate Service Provider Suggestions
The Knowledge Base helps the AI know which service providers to use for specific buildings or issue types. Instead of suggesting any electrician, it knows that Building 3 uses your in-house maintenance team while Building 5 requires external certified contractors.
Appropriate Communication Tone
Your company might prefer formal professional tone for all communications, while another company uses a warmer tone for residential tenants. The Knowledge Base ensures email drafts match your standards.
Workflow Ordering
The AI learns your procedures for handling cases - whether you require approvals before scheduling work, what information is needed in work orders, and how to structure ERP tickets according to your format.
Team Customization Examples
Different companies have different approaches, and the Knowledge Base respects these preferences:
Issue Detection Preferences:
Team A configured their Knowledge Base to split every distinct issue into separate cases, even if mentioned in the same email. When a tenant reports "heating AND elevator problems," the AI creates 2 separate cases.
Team B configured their Knowledge Base to group related issues together unless they affect different buildings. The same email creates 1 case with both issues tracked together.
Both approaches are correct - they reflect each company's documented preference.
Service Provider Preferences:
Team A prefers their contracted electrician for all buildings except commercial properties. The Knowledge Base reflects this, so the AI consistently suggests the right provider based on building type.
Team B uses different providers for different neighborhoods. The Knowledge Base includes these location-based rules, so the AI automatically selects the appropriate provider.
Communication Standards:
Team A uses formal professional tone for all communications. Email drafts are structured and business-like.
Team B uses warm professional tone for residential tenants but formal tone for commercial property owners. The AI adapts based on the recipient.
Who Manages the Knowledge Base
Captured Knowledge implementation specialists configure and maintain your Knowledge Base. You don't manage it directly.
Why Specialists Handle This:
Technical Expertise Required
The Knowledge Base must be structured in specific ways for the AI to interpret correctly. This requires understanding of AI behavior, prompt engineering, and knowledge representation - technical skills that property managers don't need to learn.
Quality Assurance
Specialists test all changes before deployment to verify the AI interprets the knowledge correctly and suggestions improve as expected. This prevents unintended behavior changes.
System Integration
As Captured Knowledge's AI capabilities evolve, the Knowledge Base structure adapts accordingly. Specialists ensure your knowledge remains effective as the platform improves.
Consistency Standards
Specialists maintain formatting and structure standards that ensure the AI can use the knowledge effectively across all scenarios.
Initial Setup Process
When you start using Captured Knowledge, an implementation specialist works with you to build your initial Knowledge Base.
What Happens During Onboarding (typically 3-5 weeks):
Week 1-2: Discovery
Specialist interviews your team about procedures and workflows
Reviews sample communications to understand your tone and style
Documents service provider preferences and building-specific requirements
Discusses how you categorize issues and handle different scenarios
Week 2-3: Knowledge Base Development
Specialist creates knowledge content based on your procedures
Structures information for AI consumption
Tests AI behavior with sample scenarios from your actual emails
Refines based on test results
Week 4: Validation
You review AI suggestions on test cases
Provide feedback on accuracy and completeness
Specialist adjusts knowledge content based on your feedback
Final approval before going live
Week 5: Go-Live and Monitoring
Knowledge Base activated
Specialist monitors early cases closely
Makes adjustments based on real-world performance
Schedules follow-up reviews
When the Knowledge Base Needs Updating
Watch for these signs that your Knowledge Base may need updates:
Consistent Patterns of Incorrect Suggestions
AI repeatedly marks routine issues as urgent (or vice versa)
Wrong service providers suggested for specific buildings
Email tone doesn't match your standards
Missing required steps in workflows
Changed Business Circumstances
New service provider contracts
Updated communication policies
Changed emergency procedures
New building-specific requirements
Revised approval workflows
Seasonal Adjustments
Heating urgency rules for winter months
Air conditioning priorities for summer
Seasonal maintenance schedules
Immediate Corrections - What You Can Do Now
Knowledge Base updates take time (typically 3-5 days). When the AI makes incorrect suggestions, you have several immediate correction methods:
Method 1: Chat with Gustav
When to use: Refine the workflow or add context the AI missed for this specific case.
How it works:
Open the case and click the chat icon
Explain what's wrong or what context the AI needs
AI reprocesses with your feedback and shows proposed changes
Preview changes before approving
Continue conversation to refine further if needed
Example: AI suggested calling a plumber for a heating issue. You explain: "This is about the heating system maintained by our in-house staff, not a plumbing issue." AI updates the workflow accordingly.
Benefits: Conversational, shows preview, maintains context for follow-up refinements.
Method 2: Reprocess Entire Case
When to use: Case categorization is completely wrong, or you want the AI to start fresh.
How it works:
Open the case and select "Reprocess Case"
Optionally provide instructions (up to 2000 characters)
AI sees all activity history and regenerates suggestions
Changes apply immediately (no preview)
Example: Case categorized as "routine plumbing" but it's actually an urgent heating emergency affecting 12 apartments in winter. Reprocessing with instructions about urgency creates appropriate emergency workflow.
Benefits: Complete fresh analysis, good when categorization is fundamentally wrong.
Method 3: Create a New Case from an Email
When to use: Single email linked to wrong case or needs different handling.
How it works:
Navigate to Email Messages tab in the case
Find the email and click "Create new case"
Optionally provide instructions about how to handle it
AI creates a new case from just that email
Example: Tenant replied to old email thread about plumbing from last month, but this time asking about a new heating issue. Creating a new case gives the heating issue its own case instead of adding it to the old plumbing case.
Benefits: Surgical fix for specific email misassignment.
Method 4: Manual Editing
When to use: Quick precise tweaks to AI suggestions before executing actions.
What you can edit: Service providers, email content, descriptions, deadlines, case categories, any field in draft actions.
Example: AI correctly identified electrical problem but suggested Provider A. You prefer Provider B for this building. Simply change the provider dropdown and save.
Benefits: Fastest method, immediate, no AI reprocessing needed.
Choosing the Right Method
Quick decision guide:
Scenario | Method |
Wrong service provider | Manual edit |
Email linked to wrong case | Create new case |
Wrong category, workflow completely off | Reprocess case |
Workflow needs several refinements | Chat with Gustav |
AI missed building policies | Chat with Gustav |
New information just arrived | Reprocess case |
When to Request Knowledge Base Updates
Immediate corrections solve today's problem. Knowledge Base updates prevent tomorrow's.
Use Immediate Corrections When:
One-time or rare situation
Special exception to normal procedures
Temporary situation (contractor unavailable today only)
You're still figuring out the best approach
Request Knowledge Base Update When:
You're making the same correction on 5+ cases
Same issue repeats weekly or monthly
Problem affects entire building or property type
Business circumstances changed (new contracts, policies)
AI consistently misunderstands your preferences
The Rule of Three
If you correct the same issue THREE times, request a Knowledge Base update.
This prevents you from becoming a "human knowledge base" - repeatedly teaching the AI the same thing.
Example timeline:
Monday: Use Chat to tell AI "Building 3 has in-house maintenance"
Wednesday: Use Chat again for another Building 3 issue
Friday: Third time? Request Knowledge Base update
After the update, all future Building 3 cases automatically use in-house maintenance without your intervention.
How to Request Knowledge Base Updates
Contact your implementation specialist when you notice consistent patterns of incorrect suggestions affecting multiple cases.
Contact Information:
Email: [email protected]
Subject line: "Knowledge Base Update Request - [Your Company Name]"
What to Include in Your Request:
1. Specific Examples with Case Numbers
At least 3-5 case numbers where AI suggestions were wrong
Screenshots showing incorrect suggestions
Description of what was wrong
2. Expected vs. Actual Behavior
What you expected the AI to suggest
What it actually suggested
Why the suggestion was incorrect
What the correct approach should be
3. Pattern Documentation
How often is this happening? (daily, weekly, specific building only)
Which types of issues are affected?
When did this start?
How many times have you manually corrected it?
4. Current Procedures (if this represents a change)
New policies you've implemented
Updated service provider preferences
Changed approval workflows
Building-specific rules
Example Request 1: Seasonal Urgency
Subject: Knowledge Base Update Request - Riverside PropertiesHello,We need an update for heating issue categorization.Problem: AI categorizes heating issues as routine during winter months, but our policy treats all heating issues as urgent November-March.Examples: - Case #1234 (Dec 15): No heat reported, AI marked routine - Case #1567 (Dec 18): Heating system failure, AI marked routine - Case #1789 (Jan 5): Cold apartment complaint, AI marked routine - Case #1834 (Jan 12): Radiator not working, AI marked routine - Case #1891 (Jan 20): Boiler issues, AI marked routineExpected: All heating issues marked urgent November-March Actual: Being marked as routine throughout winterPattern: Affects 80% of heating cases (12 out of 15 last month). All residential buildings. I've corrected this 8 times using Chat.This aligns with Austrian tenant protection law requiring immediate response to heating failures during cold months.Could you please update the heating knowledge to reflect our seasonal urgency policy?Thank you, [Your Name] [Company Name]
Example Request 2: Service Provider Rules
Subject: Knowledge Base Update Request - Alpine PropertiesHello,We need an update for Building 3 service provider selection.Problem: AI keeps suggesting external contractors for Building 3 heating system, but we have in-house maintenance there.Examples: - Case #2001: Building 3 heating - AI suggested external plumber - Case #2034: Building 3 radiator - AI suggested heating contractor - Case #2067: Building 3 boiler check - AI suggested external serviceExpected: In-house maintenance for all Building 3 heating issues Actual: AI suggests external contractorsPattern: Happens for every Building 3 heating case (10+ examples). I've been manually changing the service provider each time.Could you please update the knowledge to specify that Building 3 uses in-house maintenance for heating system work?Thank you, [Your Name] [Company Name]
Example Request 3: Communication Style
Subject: Knowledge Base Update Request - Mountain View PropertiesHello,We need an update for work order description style.Problem: AI drafts work orders using passive constructions like "Ein Elektriker wird benötigt" instead of our active professional style.Examples showing actual AI drafts: - Case #3012: "Ein Installateur wird benötigt" - Case #3034: "Es wird ein Elektriker benötigt" - Case #3056: "Ein Heizungstechniker soll gerufen werden"Expected format (our standard): "Zur Beseitigung des [Problem] in [Location] beauftragen wir Sie mit der Durchführung der erforderlichen [Trade]-Arbeiten."Pattern: Affects all work orders (100% need manual editing). Started 2 weeks ago, have edited 23 work orders.Could you please update the work order formatting to specify our active, professional phrasing standard?Thank you, [Your Name] [Company Name]
What Happens Next:
Response Timeline:
Initial response: Within 1 business day
Implementation: 3-5 business days for straightforward updates
Complex changes: 1-2 weeks with testing
Update Process:
Specialist reviews your examples
Updates knowledge content
Tests changes with sample scenarios
Deploys to your team
Monitors first cases for accuracy
Follows up to confirm improvement
Best Practices for Working with the Knowledge Base
Provide Feedback Consistently
Note when AI suggestions are wrong and document patterns. Include enough detail for specialists to understand the issue. Share positive examples too - this helps specialists know what's working well.
Schedule Regular Reviews
Quarterly check-ins with your implementation specialist help catch issues early and proactively update for seasonal changes or business evolution.
Communicate Changes Promptly
Let your specialist know about new service provider contracts, updated policies, changed team procedures, or building-specific rule changes as soon as they happen.
Track Improvement Over Time
Notice when suggestions become more accurate. Recognize patterns the AI learns well. Share success stories with your team to build confidence in the system.
Understanding Knowledge Base Limitations
The Knowledge Base Cannot:
Replace Human Judgment
AI provides suggestions based on patterns. You make final decisions. Complex situations require human evaluation. Edge cases may not be covered.
Handle Every Unique Situation
The Knowledge Base captures general patterns. Unusual circumstances need human handling. One-time exceptions aren't in the Knowledge Base. You can always override AI suggestions.
Read Your Mind
AI works from documented knowledge. Undocumented preferences won't be reflected. You must communicate changes. Continuous feedback improves accuracy over time.
Provide Instant Perfection
The initial Knowledge Base is a starting point that improves through feedback and updates. Accuracy increases over time as the AI learns your specific patterns.
Privacy and Data Security
Your Knowledge Base is Private:
Each company has a separate Knowledge Base. Your procedures aren't shared with other customers. Service provider preferences stay confidential. Communication templates remain private.
What's in the Knowledge Base:
Policies and procedures you've shared during onboarding, service provider preferences you've documented, communication standards you've provided, and company-specific rules and requirements.
What's NOT in the Knowledge Base:
Tenant personal information, financial details, property ownership information, or email content and case details from actual cases.
Related Articles
Introduction to Captured Knowledge - Core concepts and how the system works
Email Ingestion Service - How the AI processes incoming emails using the Knowledge Base
Troubleshooting Guide - What to do when AI suggestions seem wrong
The Knowledge Base is what makes Captured Knowledge smart about your specific business. While implementation specialists manage it directly, your feedback and documentation ensure it accurately represents your company's way of working. The better the Knowledge Base, the more helpful the AI suggestions become.