Artificial Intelligence – who hasn’t heard of it?
For the past few years, we have truly lived in the world of AI innovation. And, of course, the most common places we have seen it used are in generative AI models such as ChatGPT, Gemini, Claude – the list goes on. Recently, however, the industry has seen the birth of a new model type: Agentic AI. These new models, including OpenClaw, can make autonomous decisions, opening an entirely new door across multiple industries.
What Is Generative AI?
To start, let’s get a background on the models we’re more familiar with. Generative AI is largely used for content creation, such as images, text, code, and audio, after receiving a prompt from a human user. While this in itself is a great accomplishment, it is still a reactive model– it can only respond to the instructions it is given. Beyond this, generative artificial intelligence has become highly skilled at understanding language and aiding in more creative ventures.
However, it has its flaws. For one, generative AI models are not memory-retentive. They will not carry information over between sessions, sometimes even in the same one if the interaction is prolonged. This makes them unreliable for long-term projects and jobs.
As we discussed before, a model like ChatGPT will not go beyond whatever the user asks of it. In other words, it has no autonomy. It’s a tool made for people to use and interact with, not one that will take its own initiatives. But as we will see, that may not be the case anymore.
What Is Agentic AI?
Agentic AI, in contrast to generative AI, has been designed especially for autonomy, meaning that it will be able to make its own decisions, among numerous other developments. According to the Aziro Marketing Team, agentic AI will take the large language models (LLMs) in use a step further by incorporating “goal-oriented planning, persistent memory, and execution engines” (Aziro 25). These assistive models will be able to take high-level goals and be able to make the decisions and actions to meet them, regardless of human interaction.
Agentic AI is built on a reasoning-to-action loop. This can generally be broken down into five steps:
Task decomposition – A larger goal is broken up into smaller subtasks
Delegation – These tasks are split amongst relevant tools or agents
Observation – The AI reviews the results of each section, modifying outputs or adjusting the framework accordingly
Synthesis – It relates the outputs to certain tasks to complete the goal
Adaptation – Based on results, the agent modifies its “route” and continues the loop
From a human standpoint, this reasoning process can also help users and teams relying on the AI’s task management to trust its decisions more, as it should be able to provide concrete reasoning for each of its actions.
As of right now, a major example of agentic AI is the tool OpenClaw. OpenClaw allows users to create various assistants for a multitude of desktop applications, including WhatsApp, Gmail, etc. These assistants are able to complete tasks within these apps, including sending emails, organizing schedules, and handling web research. The employment of AI will certainly propel numerous fields, including cybersecurity, healthcare, and finance.
The Dilemma of Ethics
As helpful as this type of model may seem, it certainly has its drawbacks, and at the forefront of worries at the moment is ethics. While the AI is being trained to mimic human decision-making, it may also take actions that go against common morals due to its reasoning process. A Boston University professor, Van Alstyne, shares his thoughts, questioning, “What happens when agent decision ability exceeds its formal authority?” (Murray 25). These questions of authority would be especially present in high-stakes settings, including government decisions and security.
To mediate the potential effects of unlimited control, he states that companies will have to place rules on what the agents can do, as well as create specific interfaces that they can use.
Another pressing issue is within the job market. With agentic artificial intelligence being designed specifically to take the place of human tasks, we would be faced with even more job insecurity than what we are currently experiencing with AI’s effects. Currently, around 13% of Americans say they have lost their jobs due to AI or the like, and with the introduction of agents, this is sure to increase. It will be up to the industry to be able to find a balance, allowing for the wonders of AI to develop while also protecting the careers of many.
Agentic AI represents a great step forward in the world of technology and is sure to improve upon its many capabilities. Changing from a skilled chatbot to an extraordinary assistant, this evolution will surely bring open doors in numerous industries for greater potential. With its impact reaching far and wide, it will be exciting to see where it takes us next.
Authored By Diya Borundiya and Anika Yadiki