What’s on the Horizon for AI in 2025

AI is Reaching Puberty

AI will need to mature quickly. Consider AI as pre-pubescent, still a child, but growing very quickly. As AI grows, much like a child learning and developing, it becomes more understandable and predictable, figuring out what works and what doesn't. However, it's still not completely reliable or trustworthy. Considering this analogy, there’s no wonder why organizations are unsure when they will become comfortable trusting AI’s capabilities.

Businesses require a level of “Enterpriseness” incorporated into their solutions to deem them production-ready. By enterpriseness, I mean things like governance, reporting and security. These are foundational to AI’s development but too often undervalued and overlooked.  Getting this right accelerates the adoption of a much more mature and ready-to-use workforce version of AI -- one that enterprises can trust to make key decisions, provide insight, and even help operate portions of the organization. What version of AI will we end up with after it’s grown up and ready to enter the workforce? Only time will tell. But if it doesn’t deliver on its promise, it could easily be sent back to repeat a grade, which wouldn’t be good for anyone.

Data Whipped

It’s no secret that the quality of our AI outputs is directly tied to the quality of our data inputs. We recently surveyed 1,200 IT leaders and found that most were focused on security risks at the expense of data quality, sustainability and infrastructure management.  As our dependency on AI solutions grows, so will our dependency on high-quality data.

However, data quality is just one part of the puzzle. Data silos, inconsistencies and sheer volume can all impact effectiveness.  We’ll need real-time data processing and strong data governance to meet our expectations.

Utilizing systems like data catalogs and lineage tracking will become important as the adoption of AI increases. AI and Machine Learning techniques for data cleansing will help manage massive amounts of data.

We don’t have to resolve all our data problems before starting our AI journey. A basic understanding of the data estate, built on a strong, modern data infrastructure, will make a huge difference. But requiring perfection can result in nothing happening at all.

As we move beyond simple chatbots and internal systems, accountability and explainability become more critical. High-quality data ensures AI outcomes are explainable and compliant with regulations. It builds trust among users and ensures non-negotiable transparency. This is especially critical as these systems take on a more autonomous role in our daily decision-making and enterprise operations.

Agents, Agents and More Agents

No prediction for 2025 would be complete without talking about agents – much like most AI sales pitches. Without exaggeration, one of the most exciting (and potentially disruptive) advancements in AI will be the rise of agentic AI: systems capable of acting autonomously, making decisions and executing actions with minimal human intervention. Until now, most AI has been reactive – responding to inputs, providing suggestions and performing controlled, supervised tasks. Agentic AI represents a significant leap forward, enabling systems to act as extensions of an organization, taking an active role in learning, decision-making and execution.

This shift unlocks AI’s massive potential, cutting across all industries. We’re moving from AI as a responsive tool to AI that operates independently, takes the initiative and self-optimizes. These "mini experts," built on the data and systems driving our organizations, will increasingly rely on Small Language Models (SLMs), narrowing their focus. Complex processes will be handled by a network of smaller experts, dividing and conquering based on their specialized training.

As exciting as this is, the risks are undeniable. Organizations must prioritize accountability and responsibility, building robust frameworks to govern these systems and mitigate the potential for unintended consequences. Ensuring explainability will be key to maintaining trust and understanding the decisions made by these autonomous agents.

While still in its infancy, 2025 will reveal the true potential of agentic AI and test our willingness to push the boundaries of our relationship with these powerful systems. The coming year promises to be pivotal for organizations looking to harness the power of agentic AI responsibly.

Looking Ahead to 2025 with Optimism

This list could easily include topics like quantum computing, sustainability, human-to-machine collaboration, the 3-machine problem, and even the evolution beyond generative and agentic AI to physical AI.

However, the key takeaway is clear: AI is entering a new phase of maturity and adoption. Whether delivering measurable ROI, building enterprise-ready solutions, managing data dependency, empowering new autonomous agents, or shifting towards an outcome-oriented design, the common thread is evolution. AI has moved beyond isolated experimentation and trending ideas to systems that deliver real, sustainable value.

This future won’t be without its challenges. Organizations must still balance innovation with governance to maintain trust as they scale. But for those willing to embrace this magnificent, complex future, the opportunities are vast.

AI in 2025 isn’t just about technology – it’s about reshaping how we think, interact with systems and create new ideas. This next chapter of AI will belong to those who approach it with focus, intent, and the confidence necessary to evolve.