The 6 Phases
CPMAI is a Best-practice methodology for AI Projects proposed by Cognilytica.
This methodology consists of 6 phases.
Phase 1: Business Understanding
- Assess the business situation
- Should you even use an AI in the first place?
- Is it actually solving a problem?
- Scope down the task; You can’t expect an AI to solve a complex task if it can’t even solve a simple and easy one.
Phase 2: Data Understanding
- What are the core data requirements?
- Where are the data coming from?
- How much data is available?
- [Important]: Since you‘ve scoped the task in PHASE 1, you won’t have to gather millions of data to start the iteration.
Phase 3: Data Preparation
- Create a pipeline for data engineering to be able to replicate the process for faster iteration.
- Think up methods to clean, prune, and optimize data
Phase 4: Model Development
- Modeling techniques
- Algorithm selection
- Model development and iteration
Phase 5: Model Evaluation
- Does the model actually work?
- Business KPI evaluation
- Is the model right/wrong in acceptable ways?
Phase 6: Model Operationalization
- AI projects are not done after deployment.
- Continuous model management and evaluation
- Model version control
- Monitoring performance
