Agentic AI: The Evolution of Automation

Diya Poluru '29

What is agentic AI?

Agentic Artificial Intelligence (AI) is identified as an AI system that can accomplish tasks on its own with minimal to no human supervision (Stryker, n.d.). Essentially, it’s a model of artificial intelligence that is able to adapt rapidly and make decisions in an advanced way. It doesn’t need much human help, making it self-sufficient and versatile (Stryker, n.d.). For context, agentic AI refers to systems like autonomous vehicles and other forms of AI that can make decisions, reason, plan and achieve goals without human intervention (Stryker, n.d.). How does agentic AI really work?

Agentic AI operates via the integration of many complex processes. They are built upon Large Language Models (LLMs), which are advanced artificial intelligence systems trained on huge amounts of data and can utilize certain machine learning algorithms to mimic human language and intelligence (Ectors, 2025).

First, AI fetches information from systems called Application Programming Interfaces (APIs), which are defined as “a set of rules or protocols that enables software applications to communicate with each other to exchange data, features and functionality” (Goodwin, n.d.). Essentially, this means that artificial intelligence applications can refer to these APIs to gain and relay verified information from online sources or databases. This allows AI to provide users with accurate information without requiring them to manually search for it, especially since AI can generate reasoning based on factual information (Ectors, 2025). In simple terms, the aforementioned process is how AI “knows” things and how it actually gets knowledge within the span of seconds.

There are also AI agents, which are systems whose abilities stretch beyond those of Natural Language Processing (NLP), the branch of AI that allows it to understand, learn, and mimic human language. AI agents can act autonomously, making decisions, getting information from APIs, solving problems, and carrying out planning tasks (Gutowska, n.d.). Essentially, agentic chatbots have “memory”—not exactly the same as a human’s mind, but more complex than non-agentic chatbots. This allows the agents to plan ahead, split up tasks, and “learn,” in the sense that it can adapt over time to feedback or new information (Ectors, 2025).

Now—agentic AI. Agentic AI is essentially when multiple or single AI agents work to accomplish complex tasks either alone or together (Ectors, 2025). First, agentic AI collects data from sensors, APIs, or other sources, such as the user and their inputs or other databases (Stryker, n.d.). Then, it processes that data and uses NLP and other tools to interpret the task at hand, which is crucial in helping it “understand” the information and determine what to do next (Stryker, n.d.). Next, it determines its tasks, which are sometimes given by the user, before setting strategies and planning for its next steps via preset algorithms (Stryker, n.d.). Afterwards, it uses probability models and other functions to assess the potential paths and actions it could take (Stryker, n.d.).

After choosing the best way to move forward, the AI follows through with its plan by either responding to the user or performing actions through other systems like APIs (Stryker, n.d.). Usually after it does this, the AI reflects on the outcome of the task by gathering feedback and reflecting internally, which it then uses to learn and improve future actions (Stryker, n.d.). Over time, the AI can become more advanced by improving its task performance with each reevaluation (Stryker, n.d.).

What’s the difference between agentic and generative AI?

Most AI tools nowadays actively utilize both agentic and generative AI while they are performing actions (Finn & Downie, n.d.). Starting with the basics, let’s take a widely known example. ChatGPT is based on an LLM and was initially a generative AI, but is rapidly becoming an agentic AI, as OpenAI, its creator, announced previously (OpenAI, 2025). The chatbot creates and generates ideas in the form of images and text. As of July 2025, ChatGPT has been able to use a variety of agentic AI skills to accomplish more complex tasks (OpenAI, 2025).

Essentially, generative AI can generate brand-new content, and is thus often used for creating text, images, videos, and more, based on the requests of the user (Finn & Downie, n.d.). In contrast, agentic AI is more heavily focused on making and acting upon autonomous decisions, and completing complex tasks with little-to-no human supervision (Finn & Downie, n.d.). A great analogy is to think about generative AI as the right brain: creative and imaginative (Ahuja, 2025). However, agentic AI, which is focused more on decisions, reasoning, and more analytical endeavors, is similar to the left brain: logical, structured, and analytical (Ahuja, 2025). Agentic AI can actually utilize both parts, combining the entire “brain” to create something groundbreaking and truly replicate the human mind (Ahuja, 2025).

Agentic AI Applications

Agentic AI is expected to take the world by storm because of how it can be applied to “virtually any AI use case in any real-world ecosystem” (Stryker, n.d.). Finance, stocks, autonomous vehicles, healthcare, cybersecurity, supply chain management—you name it (Stryker, n.d.). Whatever users need AI to analyze and plan, agentic AI is what’s rapidly seeming to become the best choice (Stryker, n.d.). For example, in healthcare, agentic AI can “monitor patient data, adjust treatment recommendations based on new test results, and provide real-time feedback to clinicians through chatbots” (Stryker, n.d.).

Concerns & Challenges

Agentic AI has undeniable potential, but it also poses questionable risks. For example, a significant risk is the lack of transparency when it’s making decisions (Vinky, 2024). When agentic AI arrives at a particular conclusion or decision, yet is unable to explain certain reasoning or fails to be transparent about it, it raises concerns about whether information that the AI was trained on was biased, contained errors, or is untrue and could potentially do harm (Vinky, 2024). Additionally, if agentic AI is tasked with making decisions relating to ethics, morals, healthcare, criminal justice, etc., the AI may come up with decisions that do not reflect traditional human thinking based on ethical standards because agentic AI isn’t necessarily trained to understand ethics or human values, morals, or feelings, which can be conflicting and cause unforeseen consequences (Vinky, 2024).

Moreover, if AI makes harmful or negatively impactful decisions, it can be hard to determine what or who is to blame, as well as how to address and prevent these mistakes in the future. AI can also cause unforeseen interactions and cause damage to certain systems (Vinky, 2024). For instance, using other people’s work on the internet without proper consent or citations can create big issues (OpenAI is currently being sued on this very topic), and external, unpredictable interactions between certain APIs and AI agents can escalate conflicts or cause crashes in financial markets (Vinky, 2024).

Additionally, many people are growing too dependent and reliant on AI, and agentic AI may only make matters worse (Vinky, 2024). Over time, over-reliance on AI can be extremely detrimental to human involvement and oversight, as well as personal development in human intelligence, which can cause major problems in operations or in the workforce (Vinky, 2024). There is also always a risk of security breaches concerning AI, and AI could cause vast amounts of damage if certain information is compromised (Vinky, 2024). Lastly, there’s also a huge risk of agentic AI—and AI in general—replacing actual human beings in jobs and in the workforce, which is creating a spike in unemployment rates and increasing socioeconomic inequality (Vinky, 2024).

Conclusion

Agentic AI is an advanced and revolutionary technology that is heavily focused on accomplishing tasks without human supervision. Many behind-the-scenes functions enable AI to think like a human and perform tasks at a rate unseen before. Agentic AI is truly changing the game and can be applicable in various aspects of daily life, taking humankind to the next level of technological revolution. However, agentic AI could also cause harm when misused, so it’s important to harness this tool wisely to benefit human civilization.

References

Ahuja, D. (2025, April 24). The Left Brain, Right Brain, and AI’s Next Leap. Medium. Retrieved October 8, 2025, from

https://medium.com/@deepak.ahu/the-left-brain-right-brain-and-ais-next-leap-3781723b d31e

Ectors, M. (2025, February 12). AI explained: GPTs, ChatGPT Operator, AI agents and Agentic AI. Maarten Ectors. Retrieved October 8, 2025, from https://mectors.medium.com/ai-explained-gpts-chatgpt-operator-ai-agents-and-agentic-ai bb8f9d1959cd

Finn, T., & Downie, A. (n.d.). Agentic AI vs. Generative AI. IBM. Retrieved October 16, 2025, from https://www.ibm.com/think/topics/agentic-ai-vs-generative-ai#7281538

G, V. (2024, December 19). The Risks of Agentic AI: Unintended Consequences of Autonomous Decision-Making. RPATech. Retrieved October 8, 2025, from https://www.rpatech.ai/risks-of-agentic-ai/

Goodwin, M. (n.d.). What is an API? IBM. Retrieved October 8, 2025, from https://www.ibm.com/think/topics/api#:~:text=An%20API%2C%20or%20application%20programming,the%20integrity%20of%20system%20security

Gutowska, A. (n.d.). What Are AI Agents? IBM. Retrieved October 8, 2025, from https://www.ibm.com/think/topics/ai-agents

Introducing ChatGPT agent: bridging research and action. (2025, July 17). OpenAI. Retrieved October 8, 2025, from https://openai.com/index/introducing-chatgpt-agent/

Stryker, C. (n.d.). What is Agentic AI? IBM. Retrieved October 8, 2025, from https://www.ibm.com/think/topics/agentic-ai

What Are Large Language Models (LLMs)? (2025, August 15). Coursera. Retrieved October 8, 2025, from https://www.coursera.org/articles/large-language-models

What is Agentic AI? Benefits, Risks, and Outlook. (n.d.). HUMAN Security. Retrieved October 8, 2025, from

https://www.humansecurity.com/learn/topics/what-is-agentic-ai-benefits-risks-and-outlook/

What Is ChatGPT? How It Works, How to Use It, and More. (2025, August 15). Coursera. Retrieved October 8, 2025, from https://www.coursera.org/articles/chatgpt

What is Natural Language Processing? - NLP Explained. (n.d.). AWS. Retrieved October 8, 2025, from https://aws.amazon.com/what-is/nlp/