KDA Capabilities

Phase 2: DECIDE

Strategic Evaluation

Filter the market down to 2-3 viable options based on the constraints from Phase 1. Choose the optimal AI tool based on security requirements, effectiveness, and implementation complexity.

The Goal

Prevent "Over-Tooling" by selecting the "Minimum Viable Tool" that gets the job done, rather than the most hyped tool. Compare Complexity vs. Effectiveness.

Model Architecture Selection

Public/SaaS Models

Convenience of Software-as-a-Service. Best for public data and rapid prototyping.

Examples:
  • • ChatGPT Enterprise
  • • Claude (Anthropic)
  • • Google Gemini
Best For:
Public data, low-security requirements, quick deployment
Considerations:
Data privacy, vendor lock-in, API costs

Private/Open-Source Models

Control of open-source hosting. Best for sensitive data and long-term control.

Examples:
  • • Llama 3 on AWS
  • • Mistral on Azure
  • • Self-hosted models
Best For:
Sensitive data, compliance requirements, cost control
Considerations:
Infrastructure costs, maintenance overhead, expertise required

Capability Matching

Does the tool specialize in the specific modality needed? Avoid "Swiss Army Knife" tools if a specialized tool performs better for the specific workflow.

Code Specialization

Tools optimized for software development tasks.

Code generation
Debugging assistance
Code review
Examples:
GitHub Copilot, Cursor, Codeium

Text Specialization

Tools optimized for writing, analysis, and text processing.

Document generation
Summarization
Content analysis
Examples:
Claude, GPT-4, Perplexity

Image Specialization

Tools optimized for image generation and analysis.

Image generation
Image editing
Visual analysis
Examples:
Midjourney, DALL-E, Stable Diffusion

Integration Feasibility

API Integration

Can this tool plug into existing ERP/CRM systems via API? Evaluate the ease of integration and available connectors.

Key Questions:
  • • Does the tool have a REST API?
  • • Are there pre-built connectors?
  • • What is the API rate limit?
  • • Is webhook support available?

Engineering Lift

Does it require significant engineering lift to maintain? Consider the total cost of ownership, not just the initial setup.

Considerations:
  • • Maintenance overhead
  • • Required expertise level
  • • Update frequency
  • • Support availability

Decision Factors Summary

Public vs Private Models
Based on data sensitivity from Phase 1
Tool-Specific Strengths
Match capability to workflow need
Integration Requirements
API availability and complexity

Complexity vs. Effectiveness Matrix

The "Decide" phase forces a comparison of Complexity vs. Effectiveness. It encourages selecting the "Minimum Viable Tool" that gets the job done, rather than the most hyped tool.

High Effectiveness, Low Complexity
Ideal choice - maximum value with minimal overhead
Example: Specialized tool for specific workflow
High Effectiveness, High Complexity
Consider if ROI justifies the engineering investment
Example: Enterprise platform requiring custom integration
Low Effectiveness, High Complexity
Avoid - poor ROI, high maintenance burden
Example: Over-engineered solution for simple task
Low Effectiveness, Low Complexity
May be acceptable for low-stakes workflows
Example: Quick prototype or proof-of-concept

Continue Your Journey

Phase 1: KNOW

Understand your requirements and constraints

Review Phase

Phase 3: ACT

Implement with measurable success metrics

Next Phase

Process Visualization

Explore the complete KDA framework flow

View Diagram