Despite years of experimentation, most organizations agree: Artificial intelligence hasn’t remade the enterprise or shattered ROI goals — yet. In fact, McKinsey found that only about 1% of respondents in a recent survey believe they are at artificial intelligence maturity. 80% of respondents in a similar report failed to see a tangible ROI from generative AI.
The data above shows that AI system implementations from a few months ago were just a step toward real artificial intelligence value in enterprises. So, if yesterday’s cutting-edge generative AI is reaching its ceiling, what will mature AI systems look like, and where are we now on the roadmap?
What Makes Agentic AI Responsible? Combining Intelligence with Action
To articulate the style of AI systems that will drive real impact for IT professionals, we need to borrow vocabulary less associated with software. People need support in their jobs — especially in fields like IT — that can reason and take responsibility. In isolation, simple assessment or automation tools fail to alleviate the burden of high-volume tasks, like managing tickets, for IT professionals. When AI systems handle only one part of accountability, IT professionals must interpret actions or finish processes themselves. IT professionals instead need support that both helps them get ahead of high-volume tasks and understand where they need to focus and why. And that’s more challenging to develop than a straightforward automation or analysis tool alone. To meet this need, the future of enterprise AI will rest in the overlap between reason and action. It needs to take responsibility.
When we consider a person responsible, they exercise reason and wisdom to determine the best action in a situation. But that’s only half of the equation, as we expect a responsible person to execute on the action they recommend. This system of accountability (reason paired with action) is what makes or breaks IT teams — and it’s the precise support they need from an agentic AI tool. The rise of agentic AI delivers this function through agents that can learn, reason, and act in alignment with enterprise IT teams to perform tasks like ticket management or incident response.
The next step? These agents need to collaborate. Concepts like Google’s Agent2Agent Protocol and Anthropic’s Model Context Protocol (MCP) help AI agents work together by providing standards for connecting AI assistants. However, agentic AI orchestration in enterprises needs more nuance, detail, and governance. This is to support the complexity of enterprises.
The AI Evolution: Automation to Reasoning
Putting aside philosophy, how can an IT team practically develop and use AI technology that meets enterprise standards? This requires strengthening a technology’s ability to reason and automate in concert. Like any good engineering problem, we need to break these ideas down to their components, then understand how they fit together to create a functional system.
To support an enterprise IT team, mature AI systems must exercise hindsight, insight, and foresight within an enterprise’s data ecosystem. The ability to collect, analyze, and predict outcomes over time ensures AI’s assessments are accurate. This ability also forms the basis of AI with explainability. Agentic AI tools, especially those that specialize and coordinate across functions, make the shift from automation to intelligence possible. For IT teams across the business world, mature AI systems will likely one day look like a library of agents with the expertise, training, and decision-making autonomy to act on an enterprise’s specific needs. They will use current advances in automation, observability, and explainability. They will also use feedback from human IT professionals. Through this, they’ll become first responders for tasks like event management, incident response, and cost optimization.
A Roadmap to Agentic AI Orchestration in IT
Enterprise IT teams must build the right technical and operational foundations first. This will prepare them for the shift from simple automation to agentic orchestration.
- Enable basic automation – Identify basic functions within the IT department that can be standardized and then completed through automation.
- Unlock data observability across the enterprise – Add artificial intelligence and machine learning tools to basic automation. This will unlock data observability across the enterprise. These tools can look across the enterprise’s complex and often separated data environments. They make observations based on this information.
- Layer in reasoning tools – Once an enterprise’s AI system has evolved enough to make observations, the IT team can layer in reasoning tools to frame information with situation-aware context. This transforms an observation into a recommendation. At this point, an enterprise’s AI layer is mature enough to meaningfully deploy AIOps and agentic orchestration.
- Augment with deeper collaboration capabilities – Many IT teams committed to AI transformation sit here on the roadmap to agentic orchestration. At this point we as IT professionals must develop our own ability to engage in collaborative learning with AI systems. Natural conversational interfaces have made this more intuitive and allowed us to engage in literal dialogue with AI to understand what it suggests, why, and its recommended action. Further, conversation is not only useful to understand what an AI agent is doing, but also to help the agent learn from human oversight and decision-making through collaborative learning.
- Take responsibility and strengthen human/AI partnership – Mature agentic orchestration works with people across an enterprise IT team. It also works with agents trained to handle complex IT processes. It takes accountability in the IT ecosystem by learning, adapting, and generalizing.
And what will IT look like once mature agentic AI orchestration is in place? IT professionals will move away from the high-volume, lower-strategy tasks that often clog their days to tackle more nuanced tasks. Instead of wading through hundreds of tickets on the same incident, a person will teach a set of agents to work together to assess and solve the problem. They’ll instead act as human oversight and spend time checking an AI agent ecosystem’s decision-making and provide suggestions for improvement. Then, they can explain these to other C-suite decision makers. This helps include new efficiencies in the enterprise’s wider strategy. They’ll also become stewards of technology, thinking deeply about what responsible AI safety, fairness, and innovation requires in their organization.
The Path Forward: Building Tomorrow’s IT Today
The change from today’s experimental artificial intelligence to mature agentic orchestration will take time. However, the plan to get there is clear. Enterprise IT teams that rethink how they work, collaborate, and share responsibility with AI systems will be most prepared. They can then use AI’s true transformative power. It isn’t a quick switch for enterprise IT teams though. They need to build strong foundations for explainable AI systems to take accountability for important tasks. They should invest in observability, automation, reasoning, and AIOps capabilities now. When AI systems take responsibility and reach their full maturity potential, enterprise IT teams will be able to look beyond the day-to-day fray, confident in their management. They’ll be able to look forward and finally have the breathing room to ask, “What’s next?”
