Global AI spending will surpass $749 billion by 2028, IDC predicts. Furthermore, the research firm expects AI investments to outpace new spending for cloud and digital service providers by then.
Simultaneously, executives are transitioning from the bright-eyed AI experimentation phase into a new era, in which they increasingly expect to see tangible evidence of AI’s return on investment (ROI). Thus, although AI spending is certainly skyrocketing, it’s also becoming increasingly tactical.
These trends are impacting the AI landscape as we head into 2025. Let’s take a look at the most critical trends that AI decision-makers can expect to hear about over the next 12 months, starting with the most promising: agentic AI.
Agentic AI Will Make Significant Progress
Gartner has identified agentic AI as the No. 1 technology trend to watch in 2025. But what, exactly, can we expect AI agents to accomplish next year?
For starters, let’s define agentic AI. Unlike traditional AI, which executes commands in a linear way, agentic AI systems interact with tools and databases autonomously and in logical order to achieve more reliable and robust results. If you were to query an agentic AI system with a multi-step problem, it would parse your request in the most logical order instead of moving from the top down.
We can conceptualize the importance of this development by remembering the importance of the order of operations in solving simple mathematical problems. Executing problems in the correct order (in this case, PEMDAS) makes the difference between the correct answer and something wildly incorrect. This illustration of agentic AI’s importance is simplistic but captures the essence of why it could transform the business world: more reliable answers and, therefore, more reliable AI systems.
In the financial services industry, agentic AI might automate complex processes like loan underwriting. Here, the AI would decompose tasks into specialized subtasks, collaborate across systems and iteratively refine outputs based on feedback. This could potentially cut review cycles down significantly.
While it’s unlikely that agentic AI will reach the enterprise deployment stage next year, we’ll certainly see many exciting developments as industry frontrunners test these systems’ capabilities.
Specialized AI Models Will Overtake LLMs in Certain Applications
Until now, large language models (LLMs) have dominated AI discourse. However, smaller, niche-specific models will likely take center stage in 2025. These smaller models excel in precision and efficiency, catering to industries with unique demands like finance, healthcare, and robotics. Their lower data and computational requirements make them ideal for energy-efficient edge computing, enabling real-time applications in autonomous vehicles, retail, and more.
In more regulated industries, these models can often cater to more stringent compliance needs, helping to ensure accuracy and safety in deployment. That’s because they operate on a much more limited and industry-specific dataset. Another upside of this is the diminished processing power required by smaller models: The less data it has to parse, the cheaper a system is to operate. We’ll see this become an increasingly important competitive advantage as leaders look to quantify their AI systems’ ROI next year.
RegTech Will Become Increasingly Popular
Although promising, small language models alone cannot meet evolving compliance needs and regulations. AI systems require another layer of security to help ensure compliance and ethical operations.
This need, coupled with the growing complexity of global AI regulations, including the EU’s recent AI Act, will help drive a surge in regulatory technology (RegTech) adoption. RegTech tools can monitor AI systems, police compliance, and help mitigate risks, offering businesses a critical “safety layer” when adopting emerging technologies. As AI penetrates sensitive sectors, the demand for such governance solutions will grow exponentially.
RegTech applications will be especially instrumental in sectors like financial services, where managing data privacy, reducing transaction fraud, and automating compliance processes are paramount. In 2025, more firms will likely embrace AI-assisted governance tools that structure, control, and optimize the use of sensitive data while helping to ensure compliance with regional and global standards.