Agentic AI: The Next Frontier In Autonomous Work

Forbes

3 days ago

Carlos M. Meléndez is Vice President of Operations of Maxar Intelligence, Puerto Rico, a world-leading geospatial intelligence firm.
Given the ubiquity of AI in all facets of business and life, it's hard to believe that its practical use has only been around for several years. AI has been playing a role in how we receive information about virtually anything that is available on the internet. It is often the "service rep" we speak to when calling a technical support center or the assistant that helps us write a memo on ChatGPT.
Despite all the ways we interact with AI, there's a new strain of AI that is just now coming to the forefront and aiming to revolutionize how we work—agentic AI.
Not to be confused with AI agents—which automate simple tasks, such as helping you find the best flight for your trip and which almost always require humans in the loop—agentic AI learns to think and make decisions completely on its own, independent of humans.
While traditional deep-learning AI models rely on static datasets, making it difficult to remain current in real-world conditions, agentic AI introduces self-learning capabilities, allowing models to expand their knowledge over time without human intervention.
Agentic AI systems represent the next evolution in AI, a major transition from traditional AI models that simply respond to human prompts. Deloitte predicts that in 2025, 25% of companies that use generative AI will launch agentic AI pilots or proofs of concept, growing to 50% in 2027.
Early versions of agentic AI are being seen in many consumer applications. AI assistants in travel sectors, for example, are performing multi-step tasks from researching vacation options and feeding them back to you to actually booking arrangements. In the enterprise, agentic AI is beginning to autonomously manage email communications, schedule meetings and generate reports with minimal human intervention. It's also powering autonomous vehicles, enabling them to independently make decisions about routes, braking and more.
The Importance Of Keeping Agentic AI Honest
Given the enormous responsibility that will be placed on agentic AI solutions as they make their own decisions, ethical principles are more important now than in any other state of AI to date. The data they are trained on needs to be diverse and unbiased and tested against a strict set of ethical standards. In addition to prioritizing explainability, organizations should regularly assess agentic AI outputs to ensure they contribute positively to society and avoid causing harm.
Consider some of the critical decisions agentic AI could make:
As assistants to busy clinicians, agentic AI could assess a patient's radiology reports, blood test results and other information contained within their medical records to recommend a course of treatment.
Agentic AI models will eventually not only leverage data to identify red flags that could discount mortgage loan candidates, but it also will be able to autonomously act on those facts by denying or granting loans to applicants.
Given the complexity and split decisions that are made in the air traffic control environment, agentic AI could be used to communicate with pilots and manage traffic patterns and travel routes.
Although humans may one day explore space beyond their spacecraft, it's more likely that AI-powered robots will be the first to do so regularly. These agentic AI systems will navigate orbital terrain autonomously and make independent decisions about where to go and what data to gather.
Overcoming Agentic AI Challenges
When you consider that some of the decisions that agentic AI makes can be life-or-death or critical to the future of humankind, humans must continue to play a role to safeguard against potential errors or biases. Human oversight can help ensure that the system's outputs align with ethical standards and even laws.
The technology and infrastructure for agentic AI is here today and is capable of performing tasks on behalf of humans, yet it's still a long way off from widespread use. It may take some time for humans to trust it. The biggest factor to its adoption will be a cultural one, forcing early adopters to act as the canaries in the coal mine, paving the way for others to follow.
As the adoption of agentic AI continues on its slow and steady journey, below are four challenges that need to be overcome:
Before humans can be expected to relegate critical decisions to agentic AI systems, they need to trust that they're making fair, ethical and unbiased decisions. This means they need to consistently act responsibly, produce outcomes for the good of humankind and avoid bias or malintent.
While agentic AI promises productivity gains, it may also disrupt labor markets as capabilities expand. How we realign our labor force, train and educate the next generation of workers, and develop patterns of co-existence need to be considered.
We currently have no formal regulatory controls for the safe use of AI, so this will need to be established before it is widely used. New frameworks will need to address questions of liability, transparency, explainability and appropriate constraints on autonomous systems.
Data is the fuel to the success of all types of AI, especially agentic AI. To give AI the ability to make smart and effective decisions, it needs to have access to relevant data sources to provide context to its decisions. Open, integrated data sources need to be available across multiple systems.
Final Thoughts
The future of agentic AI points toward a world where autonomous systems increasingly perform tasks once accomplished by humans. The most successful state of agentic AI will likely look more like human-AI collaboration rather than complete automation. Yet, how we design, deploy and govern agentic AI will truly determine how they enhance human potential or whether they create new risks.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?