Data & Knowledge Integration

For technical users onboarding Patterns

Overview

Patterns AI connects directly to your data warehouse and builds a Knowledge Base (KB) to generate insights. The core principle is simple: if humans struggle to understand your data, so will the AI. Integration is typically performed by either the Patterns team or your data engineers.

💡 Key Components:

1. Data Warehouse Connection: Your source of truth

2. Knowledge Base (KB): Provides business context and rules

Data Warehouse Integration

A successful integration begins with your data warehouse, which must be properly structured and documented. We establish secure, read-only connections that allow Patterns AI to access your analytical data while maintaining security and performance. Patterns stores the results of queries in a back to your data warehouse in a separate schema.

Your data warehouse needs:

• Analytics-Ready Schemas: Data transformed for analytical queries

• Documentation: Clear column and table comments

• Metadata: Defined primary and foreign key relationships between tables

🔧 No Data Warehouse? Our team can help set one up to Patterns specifications.

Knowledge Base Components

The KB provides crucial business context through three integrated components:

1. Queries

Queries serve as the foundation of Patterns AI's analytical capabilities. They teach the system how to access, combine, and analyze your data effectively. Well-structured queries enable both precise reproduction of existing analyses and flexible adaptation for new questions.

Here's how to add queries to the KB:

You can simply copy/paste existing BI dashboard queries and common ad-hoc analyses. While it's helpful to have documentation for each query, the AI will attempt to understand your query and generate it's own documentation. You can also work with the AI to generate new KB queries by simply describing the desired query in natural language.

For best results, build your queries with modularity in mind. This approach allows Patterns AI to combine them in different ways to answer new questions. Regular testing against existing reports ensures accuracy and reliability.

2. Report Templates (beta)

Report templates are more advanced instructions that define the structure and presentation of insights. They combine multiple queries with analysis and presentation logic to create comprehensive, actionable analyses.

The following example demonstrates a Weekly Revenue Analysis Report Template:

### Revenue Snapshot
{uid: abcxyz} 

Current week revenue stands at {revenue_amount}, showing a {percentage}% {increase/decrease} from last week. 

• Notable changes: {highlight_key_changes}
• Largest revenue source: {top_revenue_stream}

### Year-over-Year Performance
{uid: 123456} 
Compared to the same week last year, we're seeing a {percentage}% {growth/decline}.

• YTD Performance: {ytd_metric}
• Key factors: {key_contributing_factors}

### Product Performance
{uid: abc123}
Our {top_category} category leads revenue this week at {category_revenue}.

• Top performer: {product_name}
• Watch list: {underperforming_products}

When implementing templates, the only thing that is required are references to the queries and instructions. Please reference queries with their uid that can be found in their URL, eg. {uid: abcxyz}

Begin with existing queries and outline specific insights, metrics, or comments the AI should look for in each queries result data. Instruct the AI what's important to include in the report, charts, markdown tables, etc. and how to structure it accordingly. Reports should guide readers through the analysis logically, from high-level insights to detailed findings. Templates should maintain consistency while remaining flexible enough to handle varying data scenarios.

3. Protocols (Beta)

Protocols embed your business logic into Patterns AI, ensuring analyses align with your organization's priorities and decision-making processes. They define how the system should interpret data, respond to conditions, and generate recommendations.

Here's a detailed protocol implementation example:

### Revenue Decline Protocol
When weekly revenue drops by more than 10%:
• Generate segment-level analysis
• Flag for management review
• List top 3 declining products

### Unusual Growth Protocol
When any segment grows by more than 25%:
• Validate data accuracy
• Analyze channel attribution
• Document growth drivers

### Seasonal Comparison Protocol
During known seasonal periods:
• Compare to previous year's same period
• Adjust for promotional calendar
• Note any market condition changes

Protocol integration requires careful consideration of your business rules and decision-making thresholds. Begin by documenting your standard operating procedures and key decision points. Convert these into quantifiable triggers and response actions that Patterns AI can execute automatically.

Your protocols should define clear hierarchies of importance and establish specific conditions for different types of responses. Regular review and updating of protocols ensures they remain aligned with your business objectives as they evolve.

Testing and Validation

A thorough testing process ensures reliable integration of all components. Begin by validating query results against your existing reports. Review template outputs with stakeholders to confirm they provide clear, actionable insights. Test protocol triggers with both normal and edge-case data to ensure appropriate responses.

Document any special cases or handling requirements discovered during testing. This documentation becomes valuable for ongoing maintenance and future updates to the system.

📚 Integration Support: Our team provides hands-on assistance throughout the integration process. We help customize each component to your specific needs while following established best practices for optimal results.