The Seven Patterns of AI

Introduction
The Seven Patterns of AI is a core idea in the PMI Cognitive Project Management for AI (PMI-CPMAI) methodology. It belongs to the domain of Identifying Business Needs and Solutions. The framework was developed by Cognilytica. It gives project teams a shared, technology-neutral vocabulary for describing what an AI solution is meant to do. Teams use it before they talk about algorithms, tools or vendors.

Why the Seven Patterns of AI Are Important
Many AI projects fail because teams start with the technology instead of the business problem. Others try to solve too much at once, or cannot clearly state what the AI will actually do. The Seven Patterns address these problems in five ways:

1. Common language: Business stakeholders, data scientists and project managers can describe AI solutions in the same terms.
2. Scoping and feasibility: Each pattern has typical data needs, risks, success measures and levels of complexity. Knowing the pattern lets you estimate effort and judge whether AI is the right fit.
3. Focus on business value: Patterns link a business need to a type of AI capability, not to a specific model.
4. Reuse of lessons learned: Projects that share a pattern share similar pitfalls and best practices.
5. Iteration control: CPMAI advises tackling one pattern at a time in small iterations. Combining many patterns in a single first iteration raises risk sharply.

What the Seven Patterns Are

1. Hyperpersonalization
This pattern uses AI to build a profile of each individual and adapt over time. Content, offers or experiences are tailored to a single person rather than to a broad segment.
Examples: personalized product recommendations, individualized learning paths, personalized medical treatment plans, tailored financial advice.
Key idea: treating each person as an individual (a segment of one).

2. Recognition
This pattern uses machine learning to identify and classify things in unstructured data such as images, video, audio, handwriting or text.
Examples: facial recognition, object detection, optical character recognition (OCR), speech-to-text, medical image diagnosis, defect detection on a manufacturing line.
Key idea: answering the question what is this? for unstructured content.

3. Conversation and Human Interaction
This pattern lets machines interact with humans through natural language, by voice or text, and sometimes through other human-centered modes.
Examples: chatbots, voice assistants, machine translation, content generation and summarization, sentiment-aware customer service agents.
Key idea: machines and humans communicating in human terms.

4. Predictive Analytics and Decisions
This pattern uses past and current data to predict future outcomes. It helps humans make better decisions; the human usually stays in the loop.
Examples: demand forecasting, predicting customer churn, credit risk scoring, predictive maintenance, price optimization, forecasting patient readmission.
Key idea: using data to forecast and support decisions.

5. Patterns and Anomalies
This pattern uses machine learning to find patterns, relationships and outliers in data. It shows which data fits an expected pattern and which does not.
Examples: fraud detection, cybersecurity intrusion detection, finding unusual transactions, spotting data entry errors, discovering hidden clusters in data.
Key idea: finding the needle in the haystack, or the hidden structure in data.

6. Goal-Driven Systems
This pattern uses methods such as reinforcement learning to find the best solution to a problem through trial and error, guided by a defined goal or reward.
Examples: game-playing AI, resource optimization, scenario simulation, real-time bidding, iterative strategy optimization, robotic learning of tasks.
Key idea: the system learns the best path to a goal by trial and error.

7. Autonomous Systems
This pattern covers physical or software systems that complete tasks, reach goals and interact with their environment with little or no human involvement.
Examples: self-driving vehicles, autonomous drones, autonomous robots in warehouses, self-managing software processes, autonomous documentation systems.
Key idea: removing the human from the loop. This is usually the most complex and highest-risk pattern.

How the Seven Patterns Work in a CPMAI Project
The patterns are applied mainly in CPMAI Phase I: Business Understanding, and they influence every later phase. A typical flow has six steps:

Step 1: Start with the business problem. State the problem without mentioning technology. For example: We lose 15 percent of customers each year and do not know which ones will leave.

Step 2: Ask whether AI is the right answer. Some problems are better solved with simple automation, rules or traditional software. CPMAI stresses avoiding AI for its own sake.

Step 3: Map the need to one or more patterns. The churn example maps to Predictive Analytics and Decisions. A system that then sends each at-risk customer a tailored retention offer adds Hyperpersonalization.

Step 4: Scope the first iteration to a single pattern where possible. Patterns can be combined; a self-driving car uses Recognition, Predictive Analytics and Autonomous Systems. However, CPMAI recommends delivering small, quick iterations that each focus on one pattern. This reduces risk and shows value early.

Step 5: Use the pattern to guide later phases. The pattern shapes the work that follows:
- Data Understanding: Recognition needs labeled unstructured data. Predictive Analytics needs historical structured data.
- Data Preparation: labeling, cleaning and feature work differ by pattern.
- Model Development: classification, regression, clustering, NLP or reinforcement learning.
- Model Evaluation: pattern-appropriate metrics, such as accuracy and precision for Recognition, or forecast error for Predictive Analytics.
- Operationalization: monitoring needs and human oversight differ. Autonomous Systems need the strongest safety controls.

Step 6: Weigh risk and trust. Patterns that remove human oversight, such as Autonomous Systems and some Goal-Driven Systems, raise more concerns about trustworthiness, safety, ethics and governance.

Quick Distinguishing Guide
Use the main purpose of the solution to pick the pattern:
- About a specific individual and tailoring to them: Hyperpersonalization
- Identifying what is in an image, sound or document: Recognition
- Talking with or understanding human language: Conversation and Human Interaction
- Forecasting a future value or outcome to support a decision: Predictive Analytics and Decisions
- Finding outliers, unusual behavior or hidden groupings: Patterns and Anomalies
- Learning the best strategy toward a goal through trial and error: Goal-Driven Systems
- Acting independently with no human in the loop: Autonomous Systems

Common Points of Confusion
- Predictive Analytics vs. Patterns and Anomalies: Predictive Analytics forecasts a future outcome. Patterns and Anomalies finds deviations or structure in existing data. Fraud detection is the classic Patterns and Anomalies example.
- Recognition vs. Conversation: Speech-to-text alone is Recognition. Understanding the request and responding in dialogue is Conversation and Human Interaction.
- Goal-Driven vs. Autonomous: Goal-Driven Systems focus on learning the best way to reach a goal, often in simulation or optimization. Autonomous Systems focus on operating independently in the real world without human intervention.
- Hyperpersonalization vs. Segmentation: Grouping customers into segments is closer to Patterns and Anomalies (clustering). Treating each person uniquely is Hyperpersonalization.

Exam Tips: Answering Questions on The Seven Patterns of AI

1. Memorize all seven names exactly. Hyperpersonalization; Recognition; Conversation and Human Interaction; Predictive Analytics and Decisions; Patterns and Anomalies; Goal-Driven Systems; Autonomous Systems. Distractor answers may use invented patterns such as Generative Pattern or Automation Pattern.

2. Find the main verb in the scenario. Identify is Recognition. Forecast or predict is Predictive Analytics. Detect unusual is Anomalies. Tailor to each user is Hyperpersonalization. Chat or translate is Conversation. Optimize by trial and error is Goal-Driven. Operate without humans is Autonomous.

3. Look for the primary pattern. When a scenario involves several patterns, the question usually asks for the one that best describes the core business objective. Pick the pattern that delivers the main value.

4. Prefer starting with one pattern. If asked how to scope an AI project or a first iteration, the best answer is usually to focus on a single pattern and deliver small, iterative wins. Building a multi-pattern system all at once is rarely correct.

5. Business first, technology second. Correct answers usually begin with understanding the business problem and then mapping it to a pattern. Be wary of options that jump straight to choosing an algorithm, buying a tool or hiring a vendor.

6. Know where patterns sit in CPMAI. Patterns are used mainly in Phase I (Business Understanding) and inform Data Understanding and the phases after it. If a question asks when to identify the AI pattern, the answer is early, during Business Understanding.

7. Link patterns to data needs. Recognition and Conversation rely heavily on large amounts of labeled unstructured data. Predictive Analytics relies on historical data. Goal-Driven Systems need a clear reward or goal and often a simulation environment.

8. Link patterns to risk. Autonomous Systems carry the highest risk and the greatest need for safety, governance and trustworthy AI controls. Questions about the pattern needing the most oversight usually point to Autonomous Systems.

9. Watch for the not-AI answer. Some scenarios describe a problem that fixed rules or traditional automation can solve. CPMAI values recognizing when AI is not needed.

10. Remove distractors in a logical order. First rule out options that do not match the data type (structured or unstructured). Next rule out those that do not match the level of human involvement. Then choose the best fit for the business goal.

Sample Practice Questions

Question 1: A bank wants to flag credit card transactions that differ significantly from a customer's normal spending behavior. Which pattern applies?
Answer: Patterns and Anomalies.

Question 2: A retailer wants to estimate next quarter's demand for each product so managers can plan inventory. Which pattern applies?
Answer: Predictive Analytics and Decisions.

Question 3: A streaming service wants every user's home screen to be uniquely arranged based on that user's viewing history and preferences. Which pattern applies?
Answer: Hyperpersonalization.

Question 4: A hospital wants software to detect tumors in MRI scans. Which pattern applies?
Answer: Recognition.

Question 5: A project team proposes building, in its first release, a system that recognizes products, chats with customers, predicts demand and restocks shelves autonomously. What should the project manager advise?
Answer: Break the project into smaller iterations, each focused on a single pattern, starting with the one that delivers the most business value at the lowest risk.

Summary
The Seven Patterns of AI give PMI-CPMAI practitioners a practical lens for turning business needs into well-scoped AI solutions. Learn each pattern's purpose, typical examples, data needs and risk profile. Remember to start with the business problem, focus on one pattern per iteration, and use the pattern to guide every later CPMAI phase. With these points in mind, you will be well prepared for exam questions on this topic.