What Will You Learn?
By the end of this lesson, you will be able to:
- Understand the complete AI Project Cycle and its six stages
- Explain the purpose of each stage in developing an AI solution
- Identify how the stages connect and flow into each other
- Apply the AI Project Cycle framework to real-world problems
Say your school wants to build an AI system that predicts which students might need extra help in mathematics before the exams. Sounds useful, right?
But where do you even begin? Do you start by collecting data? Or by writing code? Or by buying expensive computers?
Here’s the thing: building an AI solution isn’t about jumping straight into coding. It’s about following a systematic process — a cycle of steps that guides you from “I have a problem” to “I have a working AI solution.”
This process is called the AI Project Cycle.
Think of it like cooking a dish. You don’t just throw ingredients into a pan and hope for the best. You first decide what to cook (problem), gather ingredients (data), prepare them (exploration), follow a recipe (modelling), taste and adjust (evaluation), and finally serve it (deployment).
Let’s explore each step of this cycle.
What is the AI Project Cycle?
The AI Project Cycle is a structured framework that helps us develop AI solutions systematically. It consists of six stages that flow into each other:
- Problem Scoping — Define what problem you’re solving
- Data Acquisition — Collect the data you need
- Data Exploration — Understand and prepare your data
- Modelling — Build the AI model
- Evaluation — Test if the model works correctly
- Deployment — Put the solution into real-world use
These stages don’t always happen in a strict sequence. Sometimes you need to go back to an earlier stage. For example, during evaluation, you might discover you need more data — so you go back to data acquisition. That’s why it’s called a “cycle.”
💡 Key Insight
The AI Project Cycle is iterative, not linear. You may need to revisit earlier stages multiple times before your solution is ready.
The Six Stages Explained
Let’s understand each stage with a simple example: Building an AI system to detect spam emails.
Stage 1: Problem Scoping
What happens here? You clearly define the problem you want to solve.
This is the foundation of your entire project. A poorly defined problem leads to a poorly designed solution. In this stage, you ask questions like:
- What problem are we trying to solve?
- Who is affected by this problem?
- Where does this problem occur?
- Why is solving this problem important?
Tools used: The 4Ws Canvas, Problem Statement Template
Example (Spam Detection):
- What: Emails containing scams, advertisements, or malicious links
- Who: Email users who waste time sorting through junk
- Where: Email inboxes across the world
- Why: To save time and protect users from phishing attacks
Output: A clear problem statement like: “We need to automatically identify and filter spam emails so users only see legitimate messages in their inbox.”
🧪 Think About It
What happens if you skip this stage and jump directly to collecting data? You might collect the wrong data or solve the wrong problem entirely!
Stage 2: Data Acquisition
What happens here? You collect the data needed to train your AI.
AI learns from data. Without data, there’s no AI. In this stage, you:
- Identify what data you need
- Find reliable sources for that data
- Collect and store the data properly
- Ensure the data is relevant and sufficient
Key concepts: Data sources, data features, data reliability
Example (Spam Detection):
- Collect thousands of emails — both spam and legitimate (called “ham”)
- Data features might include: sender address, subject line, email body, presence of links, time sent
- Sources: Email providers, public datasets, user-reported spam
Output: A dataset ready for analysis
💡 Remember
The quality of your AI depends on the quality of your data. Garbage in = Garbage out!
Stage 3: Data Exploration
What happens here? You examine, clean, and understand your data.
Raw data is often messy. It may have missing values, errors, or irrelevant information. In this stage, you:
- Visualize the data using graphs and charts
- Identify patterns and trends
- Clean the data (remove errors, fill missing values)
- Prepare the data for modelling
Tools used: Data visualization (bar graphs, pie charts, line graphs), statistical analysis
Example (Spam Detection):
- Visualize: How many spam vs. ham emails are in the dataset?
- Pattern: Do spam emails have more links? Are they sent at certain times?
- Clean: Remove duplicate emails, fix formatting issues
Output: Clean, well-understood data ready for the AI model
Stage 4: Modelling
What happens here? You build the AI model that will solve the problem.
This is where the “intelligence” gets created. Based on your data and problem, you choose an approach:
- Rule-Based Approach: You write specific rules (e.g., “If email contains ‘lottery winner’, mark as spam”)
- Learning-Based Approach: You let the AI learn patterns from data (Machine Learning)
Key concepts: AI vs ML vs Deep Learning, Rule-Based vs Learning-Based
Example (Spam Detection):
- Rule-Based: Create rules like “Block emails with certain keywords”
- Learning-Based: Train a model on thousands of examples so it learns to recognize spam patterns itself
Output: A trained AI model ready for testing
💡 Key Difference
Rule-based systems follow instructions you write. Learning-based systems discover patterns on their own from data.
Stage 5: Evaluation
What happens here? You test whether your model actually works.
A model that works perfectly on training data might fail on new data. In this stage, you:
- Test the model with data it hasn’t seen before
- Measure accuracy and errors
- Identify what the model gets right and wrong
- Decide if the model is good enough or needs improvement
Key concepts: True Positive, False Positive, True Negative, False Negative, Accuracy
Example (Spam Detection):
- Test: Give the model 1000 new emails it hasn’t seen
- Measure: How many spam emails did it correctly identify? How many legitimate emails did it wrongly mark as spam?
- If accuracy is low, go back to earlier stages (more data, better features, different model)
Output: A validated model with known performance metrics
Stage 6: Deployment
What happens here? You put your solution into the real world.
A model sitting on a computer isn’t useful. Deployment means making the solution available to actual users. In this stage, you:
- Integrate the model into a usable system
- Monitor how it performs with real users
- Collect feedback and continue improving
- Handle edge cases and unexpected situations
Example (Spam Detection):
- Integrate the model into an email service
- Spam emails automatically go to a separate folder
- Monitor: Are users happy? Are legitimate emails getting blocked?
- Update the model as new types of spam emerge
Output: A working AI solution being used by real people
The Complete Picture
Here’s how all six stages connect:
| Stage | Question It Answers | Output |
|---|---|---|
| 1. Problem Scoping | What problem are we solving? | Clear problem statement |
| 2. Data Acquisition | What data do we need? | Collected dataset |
| 3. Data Exploration | What does the data tell us? | Clean, prepared data |
| 4. Modelling | How will AI solve this? | Trained AI model |
| 5. Evaluation | Does the model work? | Validated model with metrics |
| 6. Deployment | How do users access it? | Working real-world solution |
Why Is This Cycle Important?
The AI Project Cycle matters because:
- It prevents wasted effort — You don’t build something nobody needs
- It ensures quality — Each stage has checks before moving forward
- It’s flexible — You can go back and improve any stage
- It’s professional — Real AI companies follow similar frameworks
- It’s learnable — Breaking down AI development makes it accessible
🧪 Activity: Map a Real AI Application
Think about YouTube’s recommendation system. Try to identify what might happen at each stage:
- Problem Scoping: What problem does it solve?
- Data Acquisition: What data does it collect?
- Data Exploration: What patterns might it look for?
- Modelling: How does it predict what you’ll like?
- Evaluation: How does it know if recommendations are good?
- Deployment: How do you experience the final solution?
Real-World Case Study: Diabetic Retinopathy Detection
Let’s see the AI Project Cycle in action with a real example from India.
The Problem: Diabetic retinopathy is an eye disease that can cause blindness. India has millions of diabetic patients but not enough eye doctors to screen everyone.
The Solution: Google partnered with Aravind Eye Hospital in Tamil Nadu to build an AI system that detects diabetic retinopathy from eye scan images.
| Stage | What They Did |
|---|---|
| Problem Scoping | Defined: Detect diabetic retinopathy early to prevent blindness |
| Data Acquisition | Collected over 128,000 retinal images from hospitals |
| Data Exploration | Doctors labeled images as “healthy” or “diseased” at different severity levels |
| Modelling | Built a deep learning model to recognize disease patterns |
| Evaluation | Tested with expert ophthalmologists — AI achieved 98.6% accuracy |
| Deployment | Deployed in 71 vision centers across Tamil Nadu |
Impact: Thousands of patients who might have gone blind are now getting timely treatment, thanks to AI screening.
Quick Recap
- The AI Project Cycle has six stages: Problem Scoping, Data Acquisition, Data Exploration, Modelling, Evaluation, and Deployment.
- Problem Scoping defines what you’re solving using tools like the 4Ws Canvas.
- Data Acquisition collects relevant, quality data from reliable sources.
- Data Exploration cleans and visualizes data to find patterns.
- Modelling builds the AI solution using rule-based or learning-based approaches.
- Evaluation tests if the model works correctly using metrics like accuracy.
- Deployment puts the solution into real-world use.
- The cycle is iterative — you often go back to improve earlier stages.
- Following this cycle leads to better, more reliable AI solutions.
Previous Lesson: What is Artificial Intelligence? A Complete Guide for Class 9 Students
Next Lesson: Problem Scoping in AI: How to Use the 4Ws Canvas to Define Your AI Project
EXERCISES
A. Fill in the Blanks
- The AI Project Cycle consists of ___________________ stages.
- The first stage of the AI Project Cycle is ___________________.
- In the ___________________ stage, we collect the data needed for our AI project.
- Data Exploration helps us ___________________ and prepare data for modelling.
- The two main approaches to modelling are rule-based and ___________________.
- In the Evaluation stage, we test if the model ___________________ correctly.
- ___________________ is the stage where the AI solution is put into real-world use.
- The 4Ws Canvas is used in the ___________________ stage.
- “Garbage in, garbage out” refers to the importance of ___________________ quality.
- The AI Project Cycle is ___________________, meaning we can go back to earlier stages.
B. Multiple Choice Questions
1. How many stages are there in the AI Project Cycle?
(a) 4
(b) 5
(c) 6
(d) 7
2. Which stage comes immediately after Problem Scoping?
(a) Modelling
(b) Evaluation
(c) Data Acquisition
(d) Deployment
3. What does the 4Ws Canvas help with?
(a) Data visualization
(b) Problem definition
(c) Model training
(d) Deployment planning
4. In which stage do we clean and visualize data?
(a) Data Acquisition
(b) Data Exploration
(c) Modelling
(d) Evaluation
5. Which approach involves writing specific instructions for the AI to follow?
(a) Learning-Based
(b) Rule-Based
(c) Data-Based
(d) None of the above
6. What is tested during the Evaluation stage?
(a) Data quality
(b) Problem statement
(c) Model performance
(d) User interface
7. Which stage puts the AI solution into real-world use?
(a) Modelling
(b) Evaluation
(c) Deployment
(d) Data Exploration
8. Why is the AI Project Cycle called a “cycle”?
(a) It happens only once
(b) It can be repeated and stages revisited
(c) It goes in circles forever
(d) None of the above
9. In the Aravind Eye Hospital case study, what did the AI detect?
(a) Heart disease
(b) Diabetic retinopathy
(c) Cancer
(d) COVID-19
10. Which is NOT a stage in the AI Project Cycle?
(a) Problem Scoping
(b) Data Acquisition
(c) Programming
(d) Evaluation
C. True or False
- The AI Project Cycle always follows stages in strict order without going back. (__)
- Problem Scoping is the first stage of the AI Project Cycle. (__)
- Data Acquisition involves testing the AI model. (__)
- You can use bar graphs and pie charts during Data Exploration. (__)
- Rule-based modelling lets the AI learn patterns from data automatically. (__)
- Evaluation happens before Deployment. (__)
- “Garbage in, garbage out” means data quality affects AI quality. (__)
- The 4Ws stands for Who, What, Where, and Why. (__)
- Deployment means the AI solution is ready for real users. (__)
- The AI Project Cycle has only five stages. (__)
D. Define the Following (30-40 words each)
- AI Project Cycle
- Problem Scoping
- Data Acquisition
- Data Exploration
- Modelling (in AI)
- Evaluation (in AI)
- Deployment
E. Very Short Answer Questions (40-50 words each)
- What is the AI Project Cycle and why is it important?
- What questions does Problem Scoping answer?
- Why is data quality important in AI projects?
- What is the difference between rule-based and learning-based approaches?
- What happens during the Data Exploration stage?
- Why do we need the Evaluation stage?
- What does Deployment mean in the AI Project Cycle?
- Why is the AI Project Cycle called “iterative”?
- Name the six stages of the AI Project Cycle in order.
- Give one example of an AI application and identify its problem statement.
F. Long Answer Questions (75-100 words each)
- Explain all six stages of the AI Project Cycle with a brief description of each.
- Describe the Aravind Eye Hospital case study and how it followed the AI Project Cycle.
- Why is Problem Scoping considered the most important stage? What happens if it’s done poorly?
- Compare rule-based and learning-based approaches to modelling with examples.
- Explain what happens during Data Exploration and why it’s necessary before modelling.
- How does the Evaluation stage help improve an AI model?
- Apply the AI Project Cycle to build an AI system for detecting fake news. Describe what would happen at each stage.
ANSWER KEY
A. Fill in the Blanks – Answers
- six — The AI Project Cycle has six distinct stages from problem to deployment.
- Problem Scoping — This stage defines what problem the AI will solve.
- Data Acquisition — This stage focuses on collecting relevant data.
- clean/understand — Data Exploration involves cleaning, visualizing, and understanding data.
- learning-based — The two approaches are rule-based and learning-based.
- works — Evaluation tests whether the model performs correctly.
- Deployment — This final stage makes the solution available to users.
- Problem Scoping — The 4Ws Canvas helps define the problem clearly.
- data — Poor data quality leads to poor AI performance.
- iterative — We can revisit and improve earlier stages as needed.
B. Multiple Choice Questions – Answers
- (c) 6 — The AI Project Cycle has six stages: Problem Scoping, Data Acquisition, Data Exploration, Modelling, Evaluation, Deployment.
- (c) Data Acquisition — After defining the problem, we collect the data needed.
- (b) Problem definition — The 4Ws Canvas helps clearly define the problem.
- (b) Data Exploration — This stage involves cleaning, visualizing, and understanding data.
- (b) Rule-Based — Rule-based approach uses specific instructions written by humans.
- (c) Model performance — Evaluation tests how well the model works.
- (c) Deployment — Deployment puts the solution into real-world use.
- (b) It can be repeated and stages revisited — The cycle is iterative, allowing improvements.
- (b) Diabetic retinopathy — The AI was built to detect this eye disease.
- (c) Programming — Programming is not a separate stage in the AI Project Cycle.
C. True or False – Answers
- False — The cycle is iterative; you can go back to earlier stages.
- True — Problem Scoping is the first and foundational stage.
- False — Data Acquisition is about collecting data; testing happens in Evaluation.
- True — Data Exploration uses visualizations like bar graphs and pie charts.
- False — Learning-based learns from data; rule-based uses human-written rules.
- True — Evaluation tests the model before it’s deployed.
- True — Data quality directly affects AI performance.
- True — 4Ws stands for Who, What, Where, and Why.
- True — Deployment makes the AI available for real-world use.
- False — The AI Project Cycle has six stages, not five.
D. Definitions – Answers
1. AI Project Cycle: A structured framework of six stages (Problem Scoping, Data Acquisition, Data Exploration, Modelling, Evaluation, Deployment) that guides the systematic development of AI solutions from problem identification to real-world implementation.
2. Problem Scoping: The first stage of the AI Project Cycle where we clearly define what problem we’re solving, who is affected, where it occurs, and why solving it matters.
3. Data Acquisition: The stage where we identify, collect, and store the data needed to train our AI model, ensuring data is relevant, sufficient, and from reliable sources.
4. Data Exploration: The stage where we examine, clean, visualize, and understand our data to identify patterns, trends, and prepare it for modelling.
5. Modelling (in AI): The stage where we build the AI model that will solve the problem, choosing between rule-based approaches (human-written rules) or learning-based approaches (machine learning from data).
6. Evaluation (in AI): The stage where we test the trained model with new data to measure its accuracy, identify errors, and determine if it’s ready for deployment.
7. Deployment: The final stage where the AI solution is integrated into a real-world system, made available to users, monitored for performance, and continuously improved based on feedback.
E. Very Short Answer Questions – Answers
1. What is the AI Project Cycle and why is it important?
The AI Project Cycle is a six-stage framework for developing AI solutions systematically. It’s important because it prevents wasted effort, ensures quality through checkpoints, allows iteration, and makes AI development accessible and professional.
2. What questions does Problem Scoping answer?
Problem Scoping answers the 4Ws: What problem are we solving? Who is affected? Where does the problem occur? Why is solving it important? These questions create a clear problem statement.
3. Why is data quality important in AI projects?
Data quality is crucial because AI learns from data. Poor quality data (errors, missing values, irrelevant information) leads to poor AI performance. This is called “garbage in, garbage out.”
4. What is the difference between rule-based and learning-based approaches?
Rule-based approach uses specific instructions written by humans (e.g., “if X, then Y”). Learning-based approach lets AI discover patterns from data automatically through machine learning.
5. What happens during the Data Exploration stage?
During Data Exploration, we visualize data using graphs, identify patterns and trends, clean the data by removing errors and filling missing values, and prepare data for the modelling stage.
6. Why do we need the Evaluation stage?
Evaluation tests if the model actually works with new data it hasn’t seen. It measures accuracy, identifies errors, and helps decide if the model is ready or needs improvement.
7. What does Deployment mean in the AI Project Cycle?
Deployment means putting the AI solution into real-world use — integrating it into a system, making it available to users, monitoring performance, and continuously improving based on feedback.
8. Why is the AI Project Cycle called “iterative”?
It’s called iterative because stages can be revisited. If evaluation shows poor results, you might go back to get more data or try different modelling approaches.
9. Name the six stages of the AI Project Cycle in order.
The six stages are: (1) Problem Scoping, (2) Data Acquisition, (3) Data Exploration, (4) Modelling, (5) Evaluation, and (6) Deployment.
10. Example AI application with problem statement:
YouTube Recommendation System. Problem Statement: “We need to automatically suggest videos that users will enjoy watching, based on their viewing history and preferences, to increase engagement and user satisfaction.”
F. Long Answer Questions – Answers
1. Explain all six stages of the AI Project Cycle:
The AI Project Cycle has six stages. Problem Scoping defines what problem to solve using tools like the 4Ws Canvas. Data Acquisition collects relevant data from reliable sources. Data Exploration cleans and visualizes data to find patterns. Modelling builds the AI solution using rule-based or learning-based approaches. Evaluation tests the model’s accuracy with new data. Deployment puts the solution into real-world use with ongoing monitoring. The cycle is iterative — stages can be revisited to improve results.
2. Aravind Eye Hospital case study:
Google partnered with Aravind Eye Hospital to detect diabetic retinopathy, an eye disease causing blindness. Problem Scoping identified the need for early detection with limited doctors. Data Acquisition collected 128,000+ retinal images. Data Exploration had doctors label images by disease severity. Modelling used deep learning to recognize disease patterns. Evaluation achieved 98.6% accuracy against expert ophthalmologists. Deployment reached 71 vision centers in Tamil Nadu, helping thousands of patients receive timely treatment.
3. Why Problem Scoping is most important:
Problem Scoping is foundational because everything else depends on it. A poorly defined problem leads to collecting wrong data, building irrelevant models, and creating solutions nobody needs. It’s like starting a journey without knowing your destination — you’ll waste time and resources going nowhere useful. Good problem scoping uses the 4Ws (What, Who, Where, Why) to create clear, focused problem statements that guide all subsequent stages.
4. Comparing rule-based and learning-based approaches:
Rule-based approach uses specific instructions written by humans. Example: spam filter with rules like “if email contains ‘lottery winner’, mark as spam.” It’s predictable but requires humans to think of every rule. Learning-based approach lets AI discover patterns from data automatically. Example: showing thousands of spam/legitimate emails and letting AI learn differences. It handles complexity better but needs lots of data. Most modern AI uses learning-based approaches for flexibility.
5. Data Exploration and its necessity:
Data Exploration involves examining, cleaning, and understanding data before modelling. We visualize data using graphs to spot patterns — like discovering spam emails often have more links. We clean data by removing duplicates, fixing errors, and handling missing values. We identify what features (characteristics) are most important. This stage is essential because raw data is messy. Without exploration, we might train our model on incorrect or irrelevant information, leading to poor performance.
6. How Evaluation helps improve AI models:
Evaluation tests the model with data it hasn’t seen during training. We measure metrics like accuracy (percentage of correct predictions) and analyze errors. If a spam filter incorrectly blocks legitimate emails, evaluation reveals this. Based on results, we can iterate — collect more data, try different features, or change the modelling approach. Evaluation prevents deploying faulty models and provides evidence that the solution actually works before real users depend on it.
7. AI Project Cycle for fake news detection:
Problem Scoping: Automatically identify fake news articles to prevent misinformation spread. Data Acquisition: Collect thousands of verified real and fake news articles from fact-checking websites and reliable news sources. Data Exploration: Analyze patterns — do fake articles use more emotional language? Have unusual sources? Visualize and clean the dataset. Modelling: Use NLP-based learning approach to recognize linguistic patterns distinguishing fake from real news. Evaluation: Test with new articles, measure accuracy, check for false positives (real news marked fake). Deployment: Integrate into social media platforms to flag suspicious articles for users.
Previous Lesson: What is Artificial Intelligence? A Complete Guide for Class 9 Students
Next Lesson: Problem Scoping in AI: How to Use the 4Ws Canvas to Define Your AI Project
