Agentic AI – Ongoing Coverage of Its Impact on Businesses
Agentic AI has moved well beyond the lab to become one of the most tangible forces reshaping businesses around the world. In the coming years, this technology promises not just rapid technological advances but a genuine social transformation, redefining the way we live, work, and interact with the world around us.
And the numbers back that up: according to Gartner, by 2028, 33% of enterprise applications will include some form of agentic AI, a massive jump from less than 1% in 2024. This growth isn’t gradual — it’s a paradigm shift already happening behind the scenes at major organizations, and anyone who isn’t paying attention now could find themselves playing catch-up faster than they think.
Along with that growth, an estimated 15% of everyday decisions are expected to be made autonomously, without a human needing to press a button or sign off on every step. Sound like a lot? Maybe, but when you look at repetitive processes like order approvals, data triage, system monitoring, and even customer responses, it becomes clear that number is not only plausible — it’s already being built in real time inside companies of all sizes.
But what exactly sets Agentic AI apart from the AI tools we already know? While traditional AI follows pre-defined rules and responds to direct commands, AI agents go further: they adapt to new situations, learn from experience, and operate independently to pursue goals without human intervention. They decide, execute, and even communicate with other agents — all with minimal or zero human involvement. This capability opens massive doors for industries dealing with hazardous environments, labor shortages, or highly repetitive processes.
But of course, it’s not all smooth sailing. 🌊 Alongside the transformative potential, security and governance around these solutions are emerging as urgent topics. After all, systems that act on their own need well-defined boundaries, robust control structures, and plenty of transparency to operate responsibly within organizations. Making sure these autonomous systems run safely, transparently, and responsibly will require solid governance frameworks and extensive testing. That balance between autonomy and control is perhaps the biggest challenge businesses face today when adopting this technology.
What Makes Agentic AI Different From Everything That Came Before
When most people think of AI applied to business, what still comes to mind are chatbots answering FAQs, recommendation algorithms on e-commerce sites, or predictive analytics systems that help data teams make decisions. These technologies are useful, no doubt, but they have a clear limitation: they depend on a human to interpret the results and take action. Agentic AI breaks that dependency entirely, because it doesn’t just analyze — it acts. And that difference, which might seem small on paper, changes everything in practice.
A well-configured AI agent can receive a high-level objective — like reducing customer response time by 40% — and from there it charts its own course, identifies process bottlenecks, tests different approaches, learns from the results, and adjusts its path without needing approval at every step. It can communicate with other specialized agents, query external databases, trigger APIs, fill out forms, and even escalate situations to humans when it recognizes a specific case falls outside its scope. This chain of coordinated actions is what makes agentic AI qualitatively different from anything that came before.
Another point worth highlighting is the ability to make decisions in uncertain contexts. Traditional automation tools, like the well-known RPA (Robotic Process Automation), work great when processes are linear and predictable but stall when they hit an exception. AI agents, on the other hand, were designed to handle variability. They can interpret ambiguous situations, weigh options, and choose a path based on pre-defined goals and constraints. This makes them far more resilient and applicable in real corporate environments, where exceptions are the rule.
Where Businesses Are Already Applying This Technology
The use cases for Agentic AI in business are already quite concrete and span sectors well beyond tech. In healthcare, for example, DeVry University deployed its first AI agent in April 2025 to help current and prospective students, drawing on a decade of prior experience with classroom technology and experimentation with natural language processing bots. At Stanford Health Care, AI agents are being used to redefine oncology care, easing the administrative burden on physicians that frequently leads to burnout.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
In finance, agents are already working on real-time fraud detection, credit risk analysis, and compliance operations that previously required entire teams dedicated exclusively to those tasks. Google even introduced the Agent Payments Protocol (AP2), an open framework developed with more than 60 payments and technology companies to support secure agent-led transactions. That shows the financial ecosystem is taking this very seriously.
In the industrial sector, the applications may be even more visible. IFS, for example, added AI agent development and management capabilities to its ERP platform through the acquisition of startup TheLoops, enabling companies to design, test, deploy, monitor, and fine-tune AI agents with built-in support for versioning, compliance, and performance optimization. Factories that have adopted AI agents to manage their supply chains report significant reductions in waste and downtime, because the systems can predict equipment failures, renegotiate deadlines with suppliers, and redistribute tasks on the production line in a coordinated, autonomous way.
In retail and customer service, AI agents are evolving well beyond first-generation chatbots. Walmart, for instance, has made it clear it doesn’t plan to give up its retail crown and that its AI strategy will play a key role in that. Oracle also jumped in strong, adding pre-built agents to its Fusion Cloud CX to automate sales, service, and marketing processes. Today, you can have an agent that follows the entire customer journey — from the first touchpoint to post-sale — making decisions like applying a personalized discount, activating a retention team, or identifying the right moment to offer an upgrade. 🚀
The Big Moves From the Tech Industry
The race for leadership in the Agentic AI space is in full swing, and the biggest names in tech are placing heavy bets. Microsoft has been particularly aggressive: it launched the Microsoft Agent Framework, an open-source SDK and runtime for building, orchestrating, and deploying AI agents and multi-agent workflows. The company also unveiled Agent 365 (A365), a new control plane to help IT teams manage and secure AI systems as they connect to corporate data, combating so-called agent sprawl. On top of that, Microsoft’s new Copilot agents, like Researcher and Analyst, provide real-time transparency into how data is being analyzed and cross-referenced to reach specific results.
Google hasn’t been sitting still either. The company introduced Gemini Enterprise as the gateway to agentic AI at work, replacing the former Agentspace app. The Agent2Agent (A2A) protocol was launched to connect agents from different ecosystems to one another, and the Agent Development Kit (ADK) was released as an open-source framework on Vertex AI, making it possible to build an AI agent with fewer than 100 lines of Python code.
Salesforce updated its pitch with Agentforce 360, promising the fastest path from prototypes to production-scale agents, along with the launch of the Trusted AI Foundation as the foundational operating system for enterprise AI ecosystems. Oracle overhauled its entire Fusion Cloud Applications with the release of Fusion Agentic Applications, embedding AI agents directly into transactional business workflows to make decisions without human intervention.
On the infrastructure side, Nvidia bet on open infrastructure with the Nemotron 3 family of models and launched the AgentIQ toolkit to connect different agents and frameworks. IBM introduced Enterprise Advantage, a new consulting service to help CIOs take their agentic AI applications from experimentation to large-scale production, and watsonx Orchestrate gained AgentOps capabilities with real-time monitoring. ServiceNow launched AI Experience (AIx), a multimodal, context-aware interface for its Now Platform, and in partnership with Nvidia created the open-source Apriel model for building AI agents that learn.
Cisco went all-in on agentic AI security, launching identity and access management capabilities along with a toolkit for customers to embed security controls into AI agents. CrowdStrike made a major play with its Agentic Security Platform and Agentic Security Workforce following the $290 million acquisition of Onum. And Deloitte entered the game with Zora AI, an agentic platform with a portfolio of agents for finance, human resources, supply chain, sales, and more.
The Question of Open Standards and Protocols
One of the most significant developments in the agentic AI ecosystem is the rise of the Model Context Protocol (MCP). It’s becoming the plug-and-play standard for agentic AI applications to pull real-time data from multiple sources. Thousands of MCP servers are already available from various vendors, allowing AI assistants to connect to data and services in a standardized way.
However, as with any technology that gains traction quickly, MCP also brings significant security risks. The ease of connection it offers makes the protocol attractive to malicious actors looking to exploit weaknesses in how MCP has been deployed. The security community is paying close attention, and frameworks like the OWASP Top 10 are being adapted to address the specific risks of LLM-based chatbots and AI agents that access data and tools to execute tasks — making them infinitely more capable and potentially more dangerous.
Security and Governance: The Sticking Points That Still Stall Projects
Here’s the knot most organizations are still trying to untangle. Giving real autonomy to AI systems within a corporate environment raises questions that go far beyond the technology itself. Who is responsible when an agent makes a wrong decision that causes financial loss or negatively impacts a customer? How do you ensure the agent won’t act in ways that violate regulations like Brazil’s LGPD or Europe’s GDPR? These questions don’t have simple answers, and that’s exactly why many Agentic AI projects remain stuck in the pilot phase, unable to scale into production.
A McKinsey survey found that while 39% of organizations are experimenting with agents, only 23% have started scaling AI agents within a business function. Gartner analysts confirm that most agentic AI projects today are early-stage experiments or proofs of concept, often fueled by hype and frequently misapplied. Rising costs, fragile governance, and unrealistic expectations are forcing a reassessment. So what separates the initiatives that survive from those that quietly get shelved?
Security in these systems needs to be approached in layers. The first layer is technical: agents need well-defined scopes of action, with granular permissions that determine exactly which systems they can access, which actions they can execute, and which situations should be escalated for human review. The second layer is governance: companies need to create clear audit structures with detailed logs of every decision made by agents, so it’s possible to trace the reasoning behind each action. Without that audit trail, it becomes impossible to identify problems, correct unwanted behaviors, and demonstrate regulatory compliance.
There’s also a third dimension — less technical but equally critical: internal trust. A recent survey showed that while most CIOs and CTOs are optimistic about agentic AI, operational-level IT professionals who will actually be responsible for implementing these agents have serious doubts. Compliance teams, legal departments, and even frontline employees need to understand how agents work, what their limits are, and how human oversight remains part of the process. Companies that neglect this internal communication end up facing cultural resistance, which can be just as paralyzing as any technical vulnerability.
Agents in Cybersecurity: Allies and Threats at the Same Time
Cybersecurity is at a crossroads with agentic AI. On one side, AI agents are being deployed in real security operations centers, where they correlate logs, enrich alerts, and even execute first-line containment actions. Microsoft launched AI agents for its Security Copilot focused on phishing detection, data protection, and identity management. CrowdStrike built an entire agentic security platform to outpace adversaries using AI.
On the other side, security researchers keep adding weight to an uncomfortable truth: AI agents aren’t particularly smart and can be easily tricked into doing foolish or dangerous things. This year’s Black Hat conference had as its dominant theme the emergence of AI tools for both adversaries and cyber defenders. Agentic AI is a powerful tool that can generate massive amounts of code in the blink of an eye, find and neutralize threats — but it’s not reliable enough to operate without proper oversight, and that’s the central paradox. 🔐
The Future of Work and the Reorganization of Companies
Agentic AI isn’t just changing tools and processes — it’s starting to redefine corporate org charts themselves. Microsoft went so far as to suggest that as companies automate more and more processes using agents, traditional functions like finance, marketing, and engineering could dissolve, giving way to an era of the agent boss, where humans spend far more time delegating to and orchestrating multiple bots than doing hands-on execution.
There’s also a growing debate about the future of RPA. Some IT leaders say more powerful, autonomous AI agents will replace this two-decade-old precursor technology, while others predict AI agents and RPA will work side by side. The reality will likely be somewhere in the middle, with agents taking on more complex tasks and RPA continuing to handle simpler, well-defined automations.
Another front worth watching is the impact on SaaS and partner ecosystems. The enterprise tech landscape is entering a critical inflection point, with agentic AI transforming partner ecosystems from human-mediated integration networks into autonomous, self-orchestrating, intelligent ecosystems. There are even those who argue AI agents will devour the SaaS market as we know it — a position Microsoft CEO Satya Nadella has already voiced publicly, saying agents will replace all software.
What to Expect in the Coming Years
The trajectory of Agentic AI in business is just getting started, and the speed at which frameworks, platforms, and use cases are evolving suggests the next two or three years will be particularly intense. Major players like Microsoft, Google, Salesforce, Oracle, Nvidia, IBM, and a wave of highly specialized startups are already shipping solutions that let you create, orchestrate, and monitor AI agents with far less friction than there was just a year ago. That means the barrier to entry is dropping fast, and mid-size organizations that once considered this technology out of reach now have access to practical, affordable tools.
One of the most interesting movements gaining momentum is multi-agent systems, where different specialized agents work together to solve complex problems that no single agent could handle efficiently on its own. OpenAI has already released an experimental framework called Swarm for developing agentic AI swarms — networks of autonomous agents capable of working together on complex tasks without human intervention. Nvidia CEO Jensen Huang predicts we’ll soon see hundreds of millions of digital agents operating inside companies.
Preparing enterprise architecture is also essential. While C-suite leaders say they’re investing in agentic AI, the complex enterprise architectures of large organizations often struggle with the demands of this technology. Databricks, for example, announced the acquisition of Neon, a serverless open-source Postgres startup, precisely because agentic AI requires a new type of architecture — traditional workflows create bottlenecks that hurt speed and performance.
Deloitte predicts that in 2025, 25% of companies using generative AI will launch agentic AI pilots or proofs of concept, growing to 50% by 2027. An EY executive was even more direct: companies struggling to keep up with the arrival of AI agents should brace themselves, because even more complex agentic AI technologies are coming fast — including physical AI with robots and quantum computing.
The big challenge that will accompany this growth is making sure the expansion of autonomy keeps pace with the maturity of governance and security systems within each organization. There’s no magic formula here: every company will need to find its own pace, testing with smaller projects, learning from mistakes, and building trust incrementally. What’s clear is that the question is no longer whether Agentic AI will transform business operations, but when each organization will be ready to make that transition in a smart and sustainable way. ⚙️
