Understanding Agentic AI

·

2 min read

Cover Image for Understanding Agentic AI

Large language models have become increasingly better at various things. AI is not limited to just classification or regression but has become really great at doing things like generating text and images. It can also generate steps and programs.

Agentic AI is the next frontier in AI. Power of LLMs is used to generate steps and then execute those steps one by one. As the steps change the external environment, the agent regenerates the next steps. An example of such agentic AI is asking an agent to shop on behalf of you while finding new shoes that can match your current wardrobe.

Another example of an agent is self driving car. The car is continuously observing the world around it and then taking actions towards a specific goal. That is reaching a particular destination.

Features of Agentic AI.

  1. Agents work autonomously.
    This means agents do not need explicit instructions. Just give it a goal and it figures out what to do within the constraints.

  2. Agents are goal oriented
    Agents are given a goal to achieve after which they stop. They constantly update their state to maximize this objective function.

  3. Agents act in a dynamic environment
    If we were working in a predictable environment, agents are not needed. Agents are useful where the world is dynamic and were wide range of events could happen.

Agentic AI in finance

Agentic AI will find many applications in finance. Finance industry is by nature about dynamic environment and it is often about goal orientation. Maximizing profits, investment returns, lowering risks etc.

While this field is remarkably new, we would expect things to change rapidly in coming years. Mastercard, Visa etc. have already launched agentic AI libraries.