Real-World Applications

Where does Generative AI fit?

Generative AI is useful whenever an application needs to create, transform, summarize, explain, or interact with information. The model is usually one part of a larger application rather than the entire application.

Important: A production application normally adds rules, data, permissions, validation, monitoring, and sometimes other AI systems around the model.

Example 1: Customer support

Support teams receive repetitive questions, but a useful response still needs to be clear and specific. Generative AI can draft a response from the customer's message and approved information.

Build a simple support-reply generator

The code below uses a real model API. In a production system, the policy would normally come from a database or retrieval system rather than being typed directly into the prompt.

# Run this cell in Google Colab.
%pip install -q openai

from getpass import getpass
import os

os.environ["OPENAI_API_KEY"] = getpass("Enter your OpenAI API key: ")

from openai import OpenAI
client = OpenAI()

customer_message = "How can I return my order? It arrived yesterday."
return_policy = "Customers can return unused items within 30 days. The item must be in its original condition."

prompt = f"""Write a concise support reply.
Use only the policy below.

Policy:
{return_policy}

Customer:
{customer_message}"""

response = client.responses.create(
    model="gpt-5.6-luna",
    input=prompt,
)

print(response.output_text)

Notice the architecture: the model did not invent the return policy from nowhere. The application supplied the policy as context and asked the model to turn that information into a customer-friendly response.

Example 2: Software development

Developers can use Generative AI to explain unfamiliar code, create test cases, suggest refactoring ideas, write documentation, and help investigate errors.

# Run this cell in Google Colab.
%pip install -q openai

from getpass import getpass
import os

os.environ["OPENAI_API_KEY"] = getpass("Enter your OpenAI API key: ")

from openai import OpenAI
client = OpenAI()

code = """
def calculate_total(price, quantity):
    return price * quantity
"""

response = client.responses.create(
    model="gpt-5.6-luna",
    input=f"Explain this Python function to a beginner and give two test cases.\n\n{code}",
)

print(response.output_text)

Good practice: Ask for small, verifiable changes. Then run the code, inspect the tests, and review the result instead of blindly accepting generated code.

Example 3: Company knowledge assistants

Employees often need answers hidden inside policies, product manuals, onboarding documents, or internal guides. A language model can turn retrieved information into a natural answer.

More applications you will see in practice

Generative AI is not only text

Modern AI applications can work across multiple kinds of content. The exact capabilities depend on the model and product, but the broader idea is simple: generation can happen in different modalities.

A real application is more than one API call

A production system usually surrounds the model with application logic.

How to choose a good Generative AI use case

Do not start with “Where can we add AI?” Start with a painful, repetitive, measurable task.

Simple rule: Start narrow. Prove value with one workflow before expanding the system.

Mini-project: Build an AI support assistant

Take the support example one step further. Design a small assistant that drafts answers to common customer questions.