Agentic AI – Assistants to Autonomous Agents

29 June, 2025 | Dr. Bhavanishankar K

Imagine you want to chill and plan for a 5 days trip to Manali. You would limit your budget to 25,000 INR that includes travel, stay and sightseeing. All set? Let me add a pinch of excitement to the plan. How would it be if a software program can look up the flights, find affordable hotels, freezes the itinerary, and books everything for you? In simple terms yes, an AI agent can do all of these in no time.

Here comes the buzzword Agentic AI, that behaves more like independent agents, they just don’t respond to the commands, they make decisions, take actions, learn and adjust themselves and all these with very little or no human intervention. The agentic AI can get things done on its own.

Now a sudden thought must be coming to your mind, ChatGPT is no different than Agentic AI… Hold on… Considering the opening example, the ChatGPT can help us suggesting multiple options fitting our budget, it can even suggest hotels but the Agentic AI can explore, compare and make the best plan, books the hotels and sightseeing within the budget mentioned.

Formally, given a workflow (i.e. sequence of steps to be completed to achieve a specific goal), an agentic AI will assume responsibility, make decisions in each step of a workflow and assures the desired goal is reached. The interesting thing is that the workflow gets completed autonomously. Unlike ChatGPT agentic AI doesn’t needs prompts at every step, rather a single prompt will execute entire workflow. In other words, multiple agents will complete the various steps of the workflow autonomously to achieve the goal.

The working of Agentic AI is depicted in the diagram below

We can understand the steps in detail by seeing them through the example of 5 days trip to Manali.

Step 1: The goal/ objective is to “Plan a 5-day trip to Manali”

Step 2: The goal is divided into sub-tasks including

• Book the tickets to Manali
• Search, finalize and Reserve the accommodation
• Plan and Schedule Sightseeing activities
• Estimate the charges for Travel, Food and Sightseeing
• Prepare the final Itinerary

Step 3: Prepare a plan of execution, that includes

• Deciding the best mode of travel
• Deciding on the Hotels which is affordable and good
• Finalise the schedule for visiting best spots of interest
• Finalise the sequence and book everything

Step 4: Execute the plan, this step automatically

• Searches the options for tickets and reserves it
• Reserves the hotel
• Based on the opening and closing timing of the tourist attractions, it reserves them
• Prepares the day wise plan

Step 5: Monitor the Progress

In case the flight gets cancelled, then recheck and books the next available one.

Step 6: Report Back by displaying

• The tickets
• Hotel booking
• Schedule of sightseeing
• Budget breakdown

All these are seamlessly executed without human intervention.

Examples of Agentic AI

Although the fully autonomous, reliable and safe Agentic AI is still emerging following are the AI systems demonstrating Agentic AI capabilities

• AutoGPT
• BabyAGI
• Meta’s CICERO

Applications of Agentic AI

• Enhanced Productivity
• Any- time ON operations
• Cost Reduction
• Different Perspectives

In Summary, Agentic AI takes a bold leap towards autonomy in true sense. Though the journey has just begun, it is the future worth waiting and watching.