As said by Andrew Ng, “AI will be able to do everything a human can – may be even better”, is proving true day by day and is even more enforced by arrival of Agentic AI (AAI). Agentic AI can be called as AI for AI, Super AI or AI on Steroids, which has also featured in as a top trend for 2025 as quoted by Gartner. So, let’s try to understand a bit more about it and dig deeper into the use cases it can help with along with what ethical considerations we need to take.

What is Agentic AI?

Unlike traditional AI, Agentic AI focuses on learning from the data, gather information by interacting with its environment and make independent decisions. It not just task oriented rather more focused on outcomes and work with end goal in sight while improving based on results. The 5 main characteristics of AAI are –

1.      Goal orientated – AAI are designed with keeping end goals in mind and it tries to achieve those goals autonomously without any human intervention

2.      Gathering data – AAI just do not interact with internal data sources rather it also interacts with external data through sensors, camera vision etc to gather data and develop a robust and bigger information base

3.      Processing information - all the data is processed, and a perception is built on real-time streaming data along with historical data sets

4.      Execution – based on ever evolving perception through data processing supported with neural networks, algorithms etc the outputs or tasks completion is achieved with more accuracy and autonomy

5.      Feedback through data flywheel – AAI always learn from their mistakes and keep improving by having data feeding back into the models resulting in better accuracy every time by infusing self-learning

Difference between Generative AI (GenAI) and Agentic AI (AAI) –

Most of the folks I have interacted with have asked me 1 simple question, how AAI and different than GenAI. Below is my take on the differences (obviously, inspiration taken from multiple articles I read)

 Agentic AI in Action –

The use cases of AAI are limitless or we can say, are limited only by capability and creativity. Few of them are mentioned below –

1.      Retail Store Operation – leveraging sensor data, camera feeds and inventory data such an AAI system can autonomously figure the need to replenish the products on aisles and order the robots in the stores to act on this task. Also, it can figure out water or oil spill on the floor through computer vision and order cleaning robots with the right location coordinates to wipe the floor clean and reduce accidental hazards

2.      Healthcare – AAI can autonomously monitor and refill the prescriptions for the patients, process their medical records and analysis the outcome of the MRI scans to prepare discharge summary of a patient reducing loads on medical staff who can focus on serious issues and save more lives

3.      Autonomous Driving – with Level 4 autonomy achieved in self driving taxi’s in Phoenix, AZ or San Francisco, CA – yet lot of improvements are needed. Here, AAI can play an important role in helping autonomous vehicles taking fast, untaught and accurate decisions to save people inside and outside the car by gathering, perceiving and processing information quickly

4.      Finance and Insurance – AI has been leveraged already in processing of contracts and policy analysis, but AAI can take it one notch ahead by also taking and acting on the decisions to expedite the processes and helping in reducing load from the resources

5.      Software development lifecycle – AAI can be leveraged by developer communities to quickly boost their productivity by having AAI write, compile and test the codes thus saving lot of time. These agents can also be used to autonomously monitor and automate lot of data acquisition, processing, quality and governance processes

 Ethical Considerations –

Agentic AI models bring in lot of value by leveraging autonomous capability for decision making and independently solving problems, but at the same time provides lot of room to worry about unethical implications on human life which can result in biased out puts to loss of life. Thus, it is of high importance that we implement AI governance supported through –

1.      Principles and Governance –

•          Defined and aligned Principles (Fairness, Robustness, Explainability, Soundness, Resilience, Privacy and Sustainability)

•          Operational Governance Policies to apply Principles

•          Well aligned teams with proper responsibilities and accountabilities

2.      Policies and Controls –

•          Well defined policies aligned with definition of AI and Principles

•          Controls defined and implemented within operational processes

•          Risk management framework and control tower defined

3.      Technology Enablement –

•          Risk platform identified and implemented to apply risk management framework

•          Data governance and quality assurance tools implemented

•          Tools deployed across to enable universal access

4.      Change Management -

•          Principles, policies and best practices are communicated across

•          Audiences identified and Training are defined as per the requirements

•          Proper communication mechanism in place

 Conclusion –

Agentic AI brings in new horizon and limitless industry-based use cases with lots of value adds. But we need to keep in mind that we get onto this journey with responsibility and proper change management to ensure that technology progresses together with people and not to replace or negatively impact people’s life.