Complete Crash
Course
TOPICS TO BE COVERED TODAY
• What are AI Agents• Single AI Agent •
Learning agent
• About LLM • Multi Model AI •
Task Automation agents
• How LLM works Agent •
Coding assistant agent
• Chat gpt • Sequential Pattern •
Ecommerce and
• AI Work Flow • Hierarchical pattern
research agent
• Working • Parallel pattern •
Travel agent
• Architechture • Asynchronous •
Healthcare assistant
• Agentic AI • Hybrid Pattern agent
• How it works • Reflex Agent •
Vision agent
• Real time example • Model Based Reflex •
Real life examples
• Manus AI Agent• Coding and non coding
• Goal based Agent • Agents
• Utility based Agent • Opportunities
• Popular AI agents
For every Software as a Service (SaaS) company that ,there will be a
corresponding AI agent company
Software Developer Data Scientist Social Media
Manager
DevOps Engineer Product Analyst Sales Development Representative
UI/UX Data Analyst Account Executive
Designer
Data Engineer QA Engineer Partnerships Manager
HR & Talent Business Operations Manager
Analyst
Acquisition
Level
AI Agents
Level
AI
Level Workflows
LLM
LARGE LANGUAGE MODELS -
LLM
LARGE LANGUAGE MODELS
-Input
LLM LLM Output
Large Language Models - LLM
Despite being trained on large amount of data, they have
limited knowledge of
proprietary information like our personal information or
internal company data.
LLMs are passive - they wait for our prompt and then
respond
AI Workflows
When will I have next meeting with Mr..
Singh ? It’s Tomorrow at
10 am
Inp LLM Out
ut put
AI
Workflows
What will the weather be like
tomorrow?
Inp LLM Outpu
ut t
AI Workflows
Inpu LLM LLM Output
t (Text to
Audio)
AI = Human Decision
Maker
Workflows
Inpu LLM Output
t
Task –
Create posts for Instagram and LinkedIn based on
news
articles, and publish them every day at 8 PM
Reason
Step - 1 Step - 2 Step - 3
Compile Summarize Write Posts
Links Articles
Task - Create posts for Instagram and LinkedIn
based on news articles, and publish them every day
at 8 PM
Reasonin
g
• What is the most effective way to compile News
Articles?
• No,
Should I copy and
its Probably Pastetoword
easier Article
compile into
links to athose
word
document
articles, ? use another Tool to fetch the data.
and
Yes it makes More sense
Task - Create posts for Instagram and LinkedIn based
on news
articles, and publish them every day at 8 PM
Step - Step - Step -
Reas
1 2 3
on Compile Summarize Write
Links Articles Posts
Act
via
Tool
s
Task - Create posts for Instagram and LinkedIn
based on news
articles, and publish them every day at 8 PM
Step - Step - Step -
Reason
1 2 3
Compile Summarize Write
Links Articles Posts
AI
Workflow
Iteration
Step - Step - Step -
1 2 3
Compile Summarize Write Pos
Links Articles Posts t
AI Agent
• How do I make sure Its good?
• All the requirements of the post has
been fulfilled?
• I will ask another LLM to check the post Iterati
• Let’s repeat this until the best criteria
are met
on
Step - Step - Step -
1 2 3
Compile Summarize Articles Pos
Links t Critique Final
Write Posts
Bot Post
AI Agents
AI
Agent
Reasoning Act via Iteration
Step 1 Tools Try again… Outp
Goal
Step 2 Till I get ut
Step 3 the best
response
AI Agents
Single AI Agent Multi model AI Agent
Multi AI
Simple 2 Agent Agent
Model
Writer Editor
Agent Agent
Multi AI
Simple 2 Agent Agent
Model
Tasks
Agent 1
Tasks
Agent 2
Tasks
Multi AI
Simple 2 Agent Agent
Model
Tools
Tools Tasks
Agent 1
Tasks
Tools
Agent 2
Tasks
Sequential
Pattern
Example – AI Powered Document
Processing
Agent 1 Agent 2 Agent 3 Agent 4
Text Text Action Saving
Extracti Summarizat Item To
on ion Extraction Database
Hierarchical
Pattern
Example – Writing a report for business
decision making
Manager AI
Agent
Task
Sub Agent Sub Agent Sub Agent Sub Agent
1 2 3 4
Monitor Analyses Tracks Internal Aggregates
Market Customer Performance insights and
Trends Sentiment Metrics makes final
Hierarchical
Pattern
Example – Writing a report for business
decision making
Manager AI
Agent
Task
Final
Report
Sub Agent Sub Agent Sub Agent Sub Agent
1 2 3 4
Monitor Analyses Tracks Internal Aggregates
Market Customer Performance insights and
Trends Sentiment Metrics makes final
Hybrid Pattern
Example – An Autonomous Handles real time
Vehicle sensor fusion,
Top level AI collision avoidance,
Sub
Agent and road condition
Agents
analysis
Sub Agent
Sub Agent
1 2
Plans the overall
route and traffic
strategy
Sub Agent Sub Agent
Feedback 3 4
Loop
Parallel Agent Design
System
Example – Large Scale Data
Analysis
Asynchronous Multi Agent
ExampleSystem
– AI Powered Cyber Security Threat
Detection
Types of AI
Agents
Simple Model- Goal- Utility- Learning
Reflex Agent Based Reflex Based Based Agent
Agent Agent Agent
Simple Reflex
Agent
Ye agent current situation ke basis pe kaam karta hai. Agar
condition milti hai, toh action leta hai. No memory, no learning.
[Sensor] → [Condition-Action Rule] →
[Actuator]
Example: A vacuum robot that turns left when it hits a wall.
• Robot sees “wall” → turns left
• No memory or learning bubble
Think of it like: "If this happens → do that."
Model-Based Reflex
Agent
Ye agent current + previous knowledge (internal state) ka use karta
hai. Thoda intelligent hota hai,
thoda memory use karta hai.
[Sensor] → [Update Internal State] → [Condition-Action Rule]
→ [Actuator]
Example: A smart vacuum remembers room layout and adjusts its
route.
• Agent sees obstacle
• Checks “map” in memory
• Decides path
Now it has some memory to improve its actions.
Goal-Based
Agent
Ye agent goal oriented hota hai. Har action se pehle sochta hai, kya ye
action mujhe goal ke paas le
jaayega?
.
[Sensor] → [Internal State + Goal] →
[Search/Planning] → [Action]
Example: An AI agent given the goal “book a flight
to Goa under ₹5000” Task: "Book flight under
₹5000"
Path A:
₹6000
Path B:
₹4800
It
reason
s
before
acting.
Utility-Based Agent
Ye agent sirf goal tak nahi rukta, balki har possible outcome ka utility (success level) calculate
karta
hai. Best possible result choose karta hai.
[Sensor] → [Internal State + Goal + Utility Function] → [Best Action] → [Actuator]
Example: It chooses the flight with best timing, price, and airline rating.
Options A, B, C with scores
Agent chooses highest utility
It ranks and chooses the most useful option.
Learning Agent
Ye agent apne experiences se seekhta hai, aur time ke sath improve karta hai. Isme ek learning
component hota hai.
[Learning Element] ⇄ [Performance Element]
•Example: ↑
Agent fine-tuned ↓
with feedback. Or an AI assistant that gets
[Critic (Feedback)] [Problem Generator (Try New Things)]
better at booking trips for you.
•Brain icon + memory book + graph showing improvement
•Learns from feedback
• It evolves with experience!
Contact : 7880113112 For Data Science with AI & Data Analytics Course by The iScale
Real-World Use Cases of AI Agents
Task Automation Agents
Let’s say you’re a busy founder or freelancer juggling emails, meetings, and research.
An AI agent can:
•Check your inbox
•Summarize important mails
•Schedule meetings on your calendar
•Even reply to emails on your behalf!
Tools like SuperAGI and AutoGPT can automate such workflows using multiple tools in
sequence."
Agent → reads Gmail → checks Google Calendar → sends email → done
Real-World Use Cases of AI Agents
Coding Assistant Agent
AI agents can also help developers by:
•Understanding what you want to build
•Breaking the task into steps
•Writing code
•Testing it
•Debugging errors — automatically!
Example:
Build a weather app” → agent generates frontend, backend, API call, and deploys
it!
•Text prompt → Agent writes HTML + CSS + JS
•Small browser showing output app
Real-World Use Cases of AI Agents
E-commerce & Product
Research
'Find me a smartwatch under ₹3000
with great reviews and 7-day battery.' The agent will:
•Search different e-commerce sites
•Compare ratings
•Pick the best product
•Even send you a link to buy!
This is next-gen
shopping."
• Agent surfing Amazon +
Flipkart
• Final card showing best
product + “Buy Now”
button
Real-World Use Cases of AI Agents
Healthcare Assistant Agent
In healthcare, agents are helping:
•Doctors summarize patient histories
•Suggest diagnoses based on symptoms
•Monitor patient health data
They’re like AI nurses or assistants, improving efficiency without replacing
humans.
•Patient record → AI agent → Diagnosis suggestion + report summary
•Label: Your AI Co-Doctor
Real-World Use Cases of AI Agents
Personal Travel
"Planning a trip?
Agent
You can ask your agent:
'Plan a 5-day trip to Japan with a daily
itinerary, budget under ₹80,000.'
The agent:
•Checks flights, hotels, experiences
•Plans your days
•Books everything if integrated with APIs!"
•Text prompt → AI → Day-wise itinerary + total
cost
•Option: “Download PDF / Book Now”
Popular AI Agents