Kalyan Kumar is Chief Product Officer at HCLSoftware.
If 2023 was the year of GenAI, when AI-driven tools experienced unprecedented adoption, 2025 could perhaps be the banner year for agentic AI. In mid-2024, when many organizations were still in the throes of their early infatuation with GenAI, McKinsey reported that agentic AI was on the rise, calling agents “the next frontier of generative AI.”
Despite the hype, these tools aren’t, strictly speaking, “new.” AI itself has existed for over 40 years. The major difference today is that companies have access to cheap computing power and the ability to process vast amounts of data at a much lower cost. That capability has driven the widespread adoption of AI — whether it’s machine learning, computer vision, natural language processing (NLP) or GenAI.
There’s been an ongoing conversation about human-in-the-loop systems (that is, with a human overseeing or interacting with the system), which is one of the big challenges of the industry. This is where agentic AI tools come into play. We are still working on the best ways of integrating GenAI capabilities within core AI agent platforms.
The key question is, how does that differ between consumer and enterprise use? And broadly, to what extent can AI truly take over human tasks?
Understanding The Expectation Gap
In the past, AI agents were often single-skill—think about Alexa or Siri. (“Alexa, add milk to my shopping list. Siri, tell me the weather for today.”) But now, we’re seeing a shift toward multi-skill agents that can handle more complex tasks.
Fundamentally, there are two areas of focus: consumer AI, which is the technology we interact with every day (such as Siri and Alexa or ChatGPT), and enterprise AI, which is when AI is adopted and used in business environments (which tends to involve industry-specific AI-driven platforms).
These are fundamentally different verticals, and it’s important to have different expectations for each. A big challenge in enterprise AI is that people assume business tools will work just as seamlessly as consumer AI, but that’s not always the case. There’s often an expectation gap, and we need to adjust how we introduce and integrate these technologies into enterprise settings. Once we recognize that distinction, we can better define the roadmap for agentic AI adoption in the business world.
Embracing The XDO Framework
From an enterprise standpoint, I recommend a framework called XDO to approach agentic AI implementation effectively:
• X (Experience): At its core, AI should improve human experiences—whether that’s customer experience, employee experience, partner experience or even machine-to-machine interactions in connected systems.
• D (Data): Enterprises can only make AI work if they fully understand and manage their data. The biggest challenge today is that enterprise data is often locked within applications and systems. Organizations need to separate data from applications, define metadata and structure their data catalogs, marketplaces and contracts effectively.
• O (Operations): There are two broad areas in terms of operations—IT and business.
1. IT Operations: AI agents can play a major role in automating tasks across IT operations, from detection and correction to fulfilling requests and deploying resources. Since humans aren’t naturally adept at interacting with machine data, AI agents can bridge that gap and generate meaningful insights.
2. Business Operations: Agentic AI can empower businesses with autonomous, intelligent operations, driving unprecedented efficiency and agility. It would transform workflows, decisions and customer experiences, enabling proactive adaptation and strategic growth. Without this framework, agentic AI will just become another tool in the enterprise toolbox, with limited impact and value.
Importance Of Agentic Orchestration
Businesses operate under regulatory and governance frameworks, making orchestration critical. Unlike deterministic business processes, agentic systems are inherently probabilistic. Companies will soon be dealing with an increasing number of AI agents from different vendors, each built on different technologies. The challenge is not just deploying these agents but orchestrating them across the entire enterprise.
Currently, every SaaS company is pushing AI agents, and enterprises are building their own on hyperscaler platforms. But many AI orchestration solutions only focus on managing their own agents. The real challenge is enterprise-wide orchestration—connecting various subsystems and ensuring AI-driven processes work across the entire business.
Companies that embrace this XDO approach—connecting experience, data and operations—are more likely to implement agentic AI effectively.
Steps For Strategic Implementation
In addition to keeping the XDO approach in mind, there are a few key steps companies should take to develop a well-thought-out agentic AI strategy.
First, assess whether you have an overarching enterprise AI strategy to begin with—not just for agentic AI, but for AI as a whole. Many organizations get caught up in the latest AI trends while missing or not completely understanding fundamental elements like machine learning, data analytics and model governance.
The second, or what I’d even call “Step Zero,” is establishing a strong data foundation. AI is only as good as the data it learns from—if your data isn’t clean, reliable and well-structured, even the most sophisticated AI will fail. Companies need to focus on organizing their enterprise data using techniques like knowledge graphs, data catalogs and metadata management. Knowing where your data is, how it connects and ensuring it’s structured correctly is critical.
Third, it’s imperative to integrate AI into existing enterprise systems. Businesses operate across a mix of legacy systems, cloud-native platforms, SaaS applications and distributed databases. Agentic AI must seamlessly connect across these environments, especially in long-running workflows like supply chain management or customer service. Developing a universal orchestration strategy ensures AI agents can interact effectively with different systems and processes.
Once those foundational elements are in place—data, orchestration and enterprise integration—you can start developing modular AI components. These could be GenAI features embedded in existing applications or standalone agentic tools.
Some Final Thoughts
Companies should resist the urge to chase every “shiny new toy” in AI. Instead of replacing everything, they should find ways to enhance existing tools with AI while ensuring responsible governance, monitoring and ethical use.
Businesses need to be clear about who they’re building for and what experiences they’re creating. Whether AI is enhancing CRM, customer service or supply chain operations, companies should focus on where AI fits into existing workflows and how it improves real-world outcomes.
They also need continuous monitoring and measurement, ensuring AI remains aligned with business goals. Ethical AI and governance, for instance, should be baked into the data foundation; responsible AI depends on how well data is managed.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
