Rodrigo Madanes is Global Innovation AI Officer at EY.
Businesses this year continued their measured adoption of generative artificial intelligence and experimentation with ways of fitting the technology into processes and workflows. As they make further progress on this front in 2025, though, they’ll also need to prepare for the next big wave of AI innovation expected to break: agentic AI.
Agentic AI represents a dramatic advancement in AI technology that allows AI software agents to function almost autonomously to solve complex problems. Agentic AI technology sits atop a large language model (LLM) and uses nearly human levels of cognition to sift and analyze data for continuous learning and improvement in ever-changing environments.
This new technology’s capacity to reason its way through undefined tasks and processes with minimal need for human oversight will arm businesses with more powerful analytical abilities and functionality. In the process, agentic AI will revolutionize and redefine, yet again, the working relationships between people and computers.
What Agentic AI Can Do
Agentic AI harnesses almost self-reflective levels of reasoning to deliver much higher quality outputs than GenAI can produce.
For example, users can ask an agentic AI agent to outline and write a blog article or piece of sales literature that fits a brand’s content marketing plan, take them through the logical steps of the drafting process, including editing and proofreading, and then tailor the piece for a targeted audience. For its part, GenAI without agentic capabilities would not produce the same quality of output or perform such reflective, chain-of-thought reasoning.
Furthermore, agentic AI can work with tools to perform users’ tasks. Providing agentic AI with user credentials, the LLM can access enterprise systems, browse the web and make requested purchases, and send and receive emails, among many other assignments.
Users can also direct agentic AI to interact and collaborate with other agents to complete higher-level goals. In this instance, the technology can interface with back-end systems such as CRMs or ERPs to, for example, track an order and respond to a user about its delivery status.
Integration Options
Agentic AI solutions come in two flavors for enterprises: buy or build. Generally speaking, companies can either buy infrastructure software to use or they can purchase the APIs and build their application.
With the first option, businesses can use an agentic framework platform to handle some of their more horizontal tasks. For example, a speech-to-speech agent using real-time interface technology can manage customer service calls, functioning as a contact center that’s integrated into the company’s back-end systems.
Alternatively, companies with workflows created specifically for their operations can do more custom development on agentic AI. These firms can buy a platform-building block of agents and configure them to fit their particular workflows, fashioning agents to support the way their people work, introduce efficiencies and automate some processes and tasks.
Start Preparing Now
Companies that need to build their own agentic AI solution can start by choosing a platform. As they consider their options, firms should think less about the costs they’ll absorb to integrate and test the technology and more about the commitment they’re making to an agentic AI architecture. Most agentic frameworks are cloud-based and don’t include exorbitant licensing costs. But companies will find it difficult to move to another platform after they’ve committed to one—and have started building on, writing code for and developing expertise in it.
Next, as these systems are more powerful than most currently in use, businesses will also need to get their compliance and risk management in order. This entails, among other things, ensuring that security needs are met, data sovereignty and compliance are in order and that the use of AI systems is tracked and managed according to the latest AI regulations.
Firms should also take time to experiment and develop some proofs of concept to learn what is and is not possible. Finally, they’ll need to prioritize use cases and determine the easiest and most impactful ones to implement.
Businesses this coming year will continue to labor to develop new and viable ways to implement, test, scale up and generate efficiencies and near-term value from GenAI. But they should also look to devote the ample time and resources needed to onboard and explore the many dynamic capabilities of its more powerful cousin: agentic AI.
This emerging technology promises to turbo-charge an already transformative field as agents offer the potential to expand the possibilities of collaborative work between people and computers to exponential degrees. Imbuing AI technology with the capacity for advanced cognition and autonomous problem-solving will give companies the opportunity to reimagine processes across their organizations.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?