CPMAI Phase I: Business Understanding

CPMAI Phase I: Business Understanding

Introduction
The Cognitive Project Management for AI (CPMAI) methodology is a data-centric, iterative framework for running Artificial Intelligence (AI), Machine Learning (ML) and advanced analytics projects. It builds on CRISP-DM (Cross-Industry Standard Process for Data Mining) and adds agile practices and AI-specific concerns. CPMAI has six phases:
1. Business Understanding
2. Data Understanding
3. Data Preparation
4. Model Development
5. Model Evaluation
6. Model Operationalization

Phase I, Business Understanding, is the foundation for everything that follows. In the PMI-CPMAI exam it sits inside the domain Identify Business Needs and Solutions. It is one of the most heavily tested areas because most AI project failures trace back to decisions made, or skipped, here.

Why Phase I Is Important

1. Most AI projects fail for business reasons, not technical ones
Industry research repeatedly shows that a large share of AI projects never reach production or never deliver value. Common root causes include:
• No clearly defined business problem
• Unrealistic expectations about what AI can do
• No measurable success criteria
• Insufficient or inappropriate data
• Lack of stakeholder buy-in
• Solving a problem that did not need AI at all
Phase I is designed to prevent each of these failure modes.

2. It answers the most important question: should we use AI at all?
CPMAI stresses that AI is not always the right answer. Many problems are better solved with rules-based automation, traditional software, process redesign or simple analytics. Phase I forces the team to justify AI before investing in it.

3. It aligns AI with strategy and ROI
AI projects are costly and uncertain. Phase I ties the initiative to organizational goals and defines how value (ROI) will be measured.

4. It sets the scope for iteration
CPMAI favours small, fast iterations that deliver real value. Phase I defines the smallest valuable first iteration, often called the Minimum Viable Product (MVP) or a narrowly scoped pilot. This avoids "boiling the ocean".

5. It surfaces risks early
Ethical, legal, regulatory, privacy, bias and transparency risks are much cheaper to address at the start than after a model is built.

What Phase I Is

Business Understanding is the phase in which the project team works with stakeholders to:
• Define the business problem or opportunity
• Decide whether AI is the right solution
• Identify which AI pattern(s) apply
• Set measurable business and technical success criteria
• Assess, at a high level, whether data exists to support the solution
• Identify risks, constraints, ethical issues and trustworthy-AI requirements
• Scope the first iteration
• Make a Go/No-Go decision on whether to proceed

The core idea is to begin with the business problem, not the technology. CPMAI warns against "AI for AI's sake" and against starting with a cool algorithm and searching for a problem to fit it.

How Phase I Works: Key Activities

Step 1: Define the business problem
• State the problem in business terms, not technical terms. For example, "Reduce customer churn by 10% within 12 months" rather than "Build a neural network".
• Identify who has the problem, how it is solved today, and what it costs.
• Understand the current process (the as-is state) and the desired future state (the to-be state).

Step 2: Determine whether AI is needed (the AI Go/No-Go assessment)
Ask questions such as:
• Does the problem need a system that learns from data, or can fixed rules solve it?
• Is the problem too complex or variable for rules-based programming?
• Is there tolerance for probabilistic answers? AI is never 100% accurate.
• Would a simpler non-AI solution work just as well?
If a straightforward rules-based or conventional solution works, CPMAI says do not use AI.

Step 3: Identify the AI pattern(s)
CPMAI uses the Seven Patterns of AI to classify what kind of AI solution is needed:
1. Hyperpersonalization: treating each individual as unique, as in personalized recommendations, content or offers
2. Recognition: identifying and understanding unstructured data such as images, video, audio, handwriting or speech
3. Conversation and Human Interaction: chatbots, voice assistants, natural language generation and understanding
4. Predictive Analytics and Decision Support: forecasting and using past data to guide human decisions
5. Patterns and Anomalies: finding outliers, fraud or unusual behaviour
6. Goal-Driven Systems: learning by trial and error to reach a goal, as in reinforcement learning, game playing or optimization
7. Autonomous Systems: systems that act with minimal human involvement, such as self-driving vehicles or autonomous robots
A project may combine several patterns. The pattern shapes data needs, risk, complexity and scope. Autonomous systems are generally the hardest and riskiest.

Step 4: Define success criteria
• Business KPIs: revenue, cost reduction, time saved, customer satisfaction, error reduction
• Technical or model metrics: accuracy, precision, recall, F1 score, latency (these are refined in later phases)
• Acceptable performance thresholds: how good is "good enough"? Compare against current human or system performance.
Success must be measurable and agreed with stakeholders before development starts.

Step 5: Assess data feasibility at a high level
A detailed data review happens in Phase II, but Phase I must still ask:
• Does relevant data exist?
• Can we access it legally and practically?
• Is there enough of it, and is it labelled if supervised learning is needed?
Without data there is no AI project, so this is a critical Go/No-Go factor.

Step 6: Identify stakeholders, roles and resources
• Business owners and sponsors, subject matter experts, data scientists, data engineers, IT, legal, compliance and end users
• Skills gaps, budget, infrastructure and time constraints

Step 7: Address trustworthy and ethical AI
• Bias and fairness, transparency and explainability, privacy, security, accountability, regulatory compliance, human oversight
• Decide early whether explainability is mandatory, as in regulated lending or healthcare. This affects which models can be used.

Step 8: Scope the first iteration
• Choose the smallest valuable piece of the problem
• Plan for short iterations: think big, start small, iterate often
• Avoid scope creep and over-ambitious first releases

Step 9: Make the Go/No-Go decision
Based on business value, AI suitability, data availability, feasibility, risk and resources, decide whether to proceed, pivot, rescope or stop. Stopping a poor project early is a success in CPMAI, not a failure.

Typical Phase I Outputs
• Business problem statement
• AI Go/No-Go assessment
• Identified AI pattern(s)
• Business and technical success criteria and KPIs
• ROI or value hypothesis
• Stakeholder map and roles
• Initial risk, ethical and compliance assessment
• Initial data availability assessment
• Iteration scope (MVP or pilot definition)
• Project charter or approval to proceed

Common Pitfalls Phase I Prevents
• Starting with technology instead of the problem
• Assuming AI will be 100% accurate
• Trying to solve too much in the first iteration
• Ignoring data availability until late in the project
• Unclear ownership and no business sponsor
• Ignoring ethics, bias and regulation
• Treating AI projects like traditional application development projects. AI is data-centric and probabilistic, not functionality-centric.

Exam Tips: Answering Questions on CPMAI Phase I: Business Understanding

Tip 1: Business problem first, always
If a question describes a team eager to use a specific algorithm or tool, the correct answer usually involves going back to define the business problem and success criteria. Prefer answers that begin with "clarify the business objective" over "select the model".

Tip 2: Not every problem needs AI
When a scenario can be solved with simple rules, automation or conventional analytics, the best answer is often not to use AI. Watch for answer choices that recommend a non-AI solution when it is clearly sufficient.

Tip 3: Know the seven patterns cold
Expect scenario questions such as "A bank wants to flag unusual transactions. Which pattern applies?" The answer is Patterns and Anomalies. Practise mapping examples:
• Product recommendations: Hyperpersonalization
• Facial or image identification: Recognition
• Chatbot: Conversation and Human Interaction
• Demand forecasting: Predictive Analytics and Decision Support
• Fraud detection: Patterns and Anomalies
• Game-playing or optimizing agent: Goal-Driven Systems
• Self-driving vehicle: Autonomous Systems

Tip 4: Measurable success criteria are mandatory
Any answer that defines clear, measurable, stakeholder-agreed KPIs before development is generally preferred. Vague goals such as "improve efficiency with AI" signal a Phase I gap.

Tip 5: Start small and iterate
CPMAI favours small, fast iterations. If the options include "deploy enterprise-wide AI in one release" versus "scope a narrow pilot that delivers value quickly", choose the pilot.

Tip 6: Data feasibility is part of Go/No-Go
If a scenario shows no access to relevant data, the right action is to address data availability or rescope before proceeding, not to start modelling.

Tip 7: Stopping is a valid outcome
Questions may test whether you recognize that a No-Go decision is appropriate. Ending or pivoting an unviable project early saves resources and is good practice.

Tip 8: Consider trustworthy AI early
In regulated or sensitive contexts such as healthcare, finance, hiring or the public sector, choose answers that raise explainability, bias, privacy and compliance during Phase I, not after deployment.

Tip 9: Engage the right stakeholders
The best answers involve business owners and subject matter experts, not only data scientists. A missing sponsor or unclear ownership is a red flag to resolve in Phase I.

Tip 10: Recognize phase boundaries
Know what belongs in Phase I and what belongs later:
• Detailed data quality profiling belongs to Phase II, Data Understanding
• Cleaning and labelling belong to Phase III, Data Preparation
• Algorithm selection and training belong to Phase IV, Model Development
If a question asks what to do first, the answer is usually a Phase I activity.

Tip 11: Expect probabilistic outcomes
Answers that set realistic expectations, such as acceptable accuracy thresholds, human-in-the-loop review and handling of errors, beat answers that promise perfect results.

Tip 12: CPMAI is iterative, so you may return to Phase I
Findings in later phases, such as insufficient data or poor model performance, can send the team back to Business Understanding to rescope. Answers recognizing this loop are typically correct.

Tip 13: Elimination strategy
Eliminate options that:
• Jump straight to technology or model choice
• Ignore stakeholders or business value
• Assume AI is always the answer
• Postpone ethics or compliance
• Expand scope prematurely
Then pick the option that is business-driven, measurable, data-aware, iterative and risk-conscious.

Sample Question Walkthrough
Question: A retail executive tells the project manager, "We need AI in our stores. Competitors are using it." What should the project manager do first?
A. Hire a data science team
B. Select a computer vision platform
C. Work with stakeholders to identify a specific business problem and measurable objectives
D. Collect all available store data
Answer: C. This reflects the core principle of Phase I: define the business problem and success criteria before choosing technology, resources or data.

Key Takeaways
• Phase I is the foundation of CPMAI and prevents most AI project failures
• Start with the business problem, not the technology
• Decide whether AI is truly needed
• Classify the solution using the seven patterns of AI
• Define measurable success criteria and ROI
• Check data availability early
• Address ethics, risk and trustworthy AI from the start
• Scope small, iterate fast and be willing to say No-Go

Master these principles and you will be well prepared for exam questions on CPMAI Phase I: Business Understanding.