The Economic Impact of Generative AI
Explore the economic impact of Generative AI across industries, including productivity gains, automation, workforce transformation, and the creation of new digital economies powered by large language models and AI systems.
Retrieval-Augmented Generation (RAG) for AI Applications
Machine Learning: Introduction and Core Algorithms
Generative AI and Business
Day-to-Day Usage of Web UI LLMs
Based on Andrew Ng's Generative AI course
1. Everyday Uses of LLMs
Large Language Models (LLMs) can assist professionals in many daily tasks.
Common use cases include writing, marketing, recruiting, and programming.
Writing Assistant
Example prompt:
Rewrite the following to be suitable for a professional business report:
The marketing campaign for the new product went really well, beating our new user goal by 10%.
LLM Output:
The product launch marketing campaign was successful, exceeding our user acquisition target by 10%.
Marketing Support
Example prompt:
Help me brainstorm an email campaign to reactivate lapsed users of our SaaS product.
Example strategy:
- “We miss you” email
- Feedback request email
- Exclusive offer email
- Feature spotlight email
- Customer success story email
Recruiting
Example prompt:
Summarize the following candidate review in 50 words or fewer.
Example output:
The reviewer appreciated the candidate’s enthusiasm and knowledge of the company. However, there were some concerns regarding familiarity with specific tools. Overall, the reviewer supports hiring the candidate but recommends discussing onboarding and training needs.
Programming Assistance
Example prompt:
Write Python code to calculate the average of the last column in a dataframe.
Example solution:
import pandas as pd
# Calculate the mean of the last column
mean_value = df.iloc[:, -1].mean()
print(mean_value)
2. AI Does Not Replace Jobs — It Replaces Tasks
A key concept in AI adoption:
AI automates tasks, not entire jobs.
Most jobs consist of multiple tasks.
Example: Customer Service Representative
| Task | Generative AI Potential |
|---|---|
| Answer phone calls | Low |
| Answer chat queries | High |
| Check order status | Medium |
| Record customer interactions | High |
| Evaluate complaints | Low |
Some tasks are easier to automate than others.
3. Augmentation vs Automation
AI can either assist humans or perform tasks automatically.
Augmentation
AI helps humans perform a task.
Example:
AI suggests a response for a customer service agent who reviews and edits it.
Automation
AI performs the task completely.
Example:
Automatically transcribing and summarizing customer interaction records.
Businesses often start with augmentation before moving toward automation.
4. Evaluating AI Opportunities
Two key factors determine whether AI should be used.
Technical Feasibility
Questions to ask:
- Can AI realistically perform the task?
- Could a new graduate complete the task with clear instructions?
- Can prompting, RAG, or fine-tuning improve performance?
Business Value
Questions to ask:
- How much time is spent on the task?
- Will automation make the task faster, cheaper, or more consistent?
- Will improving the task significantly impact business results?
5. Breaking Jobs Into Tasks
Job databases like O*NET break down roles into detailed tasks.
Example tasks for customer service:
- Provide information about products
- Record customer interactions
- Resolve customer complaints
- Process payments
- Update customer accounts
Analyzing tasks helps identify automation opportunities.
6. AI Potential Across Different Jobs
Computer Programmer
| Task | AI Potential |
|---|---|
| Write code | Medium |
| Write documentation | High |
| Respond to support requests | Medium |
| Review others’ code | Low |
| Gather requirements | Low |
Lawyer
| Task | AI Potential |
|---|---|
| Draft legal documents | High |
| Interpret laws and regulations | High |
| Review evidence | Low |
| Negotiate settlements | Low |
| Represent clients in court | Low |
Landscaper
| Task | AI Potential |
|---|---|
| Maintain plants | Low |
| Transport plants | Low |
| Maintain equipment | Low |
| Communicate with clients | Medium |
| Maintain website | Low |
7. New Workflows Enabled by AI
AI changes workflows and enables new ways of working.
Example: Surgeons
Without AI:
- Research procedures
- Perform surgery
With AI:
- AI assists with research and summarization
- Surgeons focus more on patient care and surgical procedures
Example: Legal Document Review
Without AI:
- Gather information
- Review documents manually
- Provide client feedback
With AI:
- AI reviews and summarizes documents
- Humans verify outputs
Example: Marketing
Without AI:
- Write marketing copy
- Publish content
With AI:
- Generate copy
- Run A/B tests
- Analyze campaign performance
- Improve prompts
8. Teams That Build Generative AI Software
Common roles include:
Software Engineer
- Builds the application
- Integrates LLM APIs
- Uses prompting techniques
Machine Learning Engineer
- Designs AI systems
- Works with RAG and fine-tuning
Product Manager
- Defines the product vision
- Identifies valuable AI use cases
Dedicated prompt engineer roles are rare.
Small Team Setup
A generative AI project can start with:
One-person team
- Software engineer
- Machine learning engineer
- Or even a motivated individual experimenting with LLMs
Two-person team
- ML engineer + software engineer
Additional Roles
- Data Engineer → manages data pipelines
- Data Scientist → analyzes data
- Project Manager → coordinates work
- ML Researcher → develops new algorithms
9. Economic Impact of Generative AI
Research suggests that generative AI may impact higher-paid knowledge workers the most.
Areas with large potential impact include:
- Software engineering
- Marketing
- Sales
- Customer operations
- Finance
- Legal services
- Product development
These areas represent a large share of generative AI's economic impact.
10. Societal Concerns About AI
Concern 1: Bias
LLMs learn from internet data, which may contain bias.
Solutions include:
- Reinforcement Learning from Human Feedback (RLHF)
- Human evaluation of outputs
- Improved training methods
Concern 2: Job Loss
Some people worry that AI will replace entire professions.
Example prediction:
Radiologists will soon be replaced by AI.
However, a more realistic view is:
Radiologists who use AI will replace those who do not.
Concern 3: Human Extinction
Some argue that AI could pose existential risks.
However:
- These concerns are largely speculative.
- Many powerful technologies carry risk but still provide enormous benefits.
Examples include airplanes and nuclear energy.
11. Artificial General Intelligence (AGI)
AGI refers to AI systems capable of performing any intellectual task a human can perform.
Examples:
- Learning to drive a car
- Writing a PhD thesis
- Performing the full role of a knowledge worker
Current AI systems are not yet AGI.
12. Responsible AI
Responsible AI requires attention to several principles.
Fairness
Ensure AI systems do not amplify bias.
Transparency
Make AI decisions understandable.
Privacy
Protect user data.
Security
Protect AI systems from attacks.
Ethical Use
Ensure AI benefits society.
13. Responsible AI Practices
Recommended practices include:
- Encourage ethical discussions within teams
- Brainstorm potential failure scenarios
- Evaluate risks related to fairness, privacy, and security
- Include diverse perspectives in decision making
14. Course Summary
Key concepts covered:
- How generative AI works
- What LLMs can and cannot do
- Common applications such as writing and summarization
- Generative AI project lifecycle
- Technologies such as prompting, RAG, and fine-tuning
- Business applications of generative AI
- Societal impacts and responsible AI
Final Thought
Generative AI will transform how people work.
The goal is not to replace humans, but to augment human capabilities and build a more intelligent world.