Top 10 AI Trends 2025: How Agentic AI and MCP Changed IT

Artificial intelligence (primarily generative and agentic AI) and machine learning innovation in just the past two years have fundamentally changed the way we do work. AI technology and skills have developed at a faster pace in 2025 than in any previous year since the inception of generative AI. In fact, some may claim 2025 as “the year of the agent” with 83% of organizations surveyed in the 2025 Cisco AI Readiness Index saying that they had planned to deploy agentic AI systems.

Generative AI and agentic AI investments began to show real gains in adoption and productivity across the board in numerous industries and operations. According to a BCG report from 2025, Generative AI has “helped companies achieve productivity improvements of between 15% and 30%, with some aspiring to as much as 80% higher productivity.” Every day, more people across every organization are applying generative and agentic AI to their day-to-day operations to troubleshoot incidents more quickly, collaborate across teams and silos, and write quality code faster.

AI Is Driving Operational Efficiency for DevOps, IT, and Security

Specific for SecOps, ITOps, NetOps, and Observability, AI agents and generative AI tools are breaking down silos across teams and making it easier to work cross-functionally. 2025 also marked the beginning of generative software and interfaces like Cisco AI Canvas where teams can easily work together in real-time and with specific context, without any additional fluff. Post-incident reports can then be quickly generated and shared with AI and comprehensive root cause analysis can be completed faster than ever before.

With the development and widespread adoption of Model Context Protocol (MCP) in 2025 as the universal protocol for AI-native APIs, the use cases for AI continued to expand even further (check out this short YouTube video to learn more about MCP). MCP enables AI apps and agents to interact seamlessly across tools, data sources, and environments through a standardized client/server architecture. MCP drove key enhancements to SecOps, ITOps, NetOps, and Observability workflows such as human-in-the-loop agentic operations and root cause analysis. Among NetOps teams, MCP adoption accelerated automation efforts by simplifying integration complexity and enabling natural language interactions with infrastructure like SD-WAN.

Of course, there is more, but with all of that in mind, let’s dive into the top 10 ways that AI technology and skills grew in 2025.

Summarizing 10 Major AI Developments in 2025

The top 10 focuses for AI developments in 2025 were the following:

• The development of precision models for specific use cases
• The emergence of security and observability as foundations for reliable AI
• Machine data now powering AI models and Agentic Ops
• Massive implications of AI on technology infrastructure
• Challenges and innovation around sustainability
• Advances to quantum networking
• Growing partner ecosystems and AI talent training
• Popularizing MCP servers and AI infrastructure
• Even more AI skills and platforms to choose from
• The evolution of AI-driven automation and self-healing systems

Deep Dive Into the Top 10 Ways That AI Evolved Throughout 2025

1. The Development of Precision Models for Specific Use Cases

Organizations moved away from massive GPU clusters toward more targeted, efficient AI solutions. Due to high costs and energy consumption of large language model training, more businesses embraced smaller, specialized models (such as Cisco’s Foundation Security Reasoning Model) that are optimized for specific business outcomes and can improve control, compliance, and efficiency.

2. The Emergence of Security and Observability as Foundations for Reliable AI

Throughout 2025, AI vulnerabilities and reliability concerns became increasingly recognized as data problems, leading to more comprehensive observability practices across the entire AI data lifecycle. Unified visibility and integrated security operations became critical for detecting threats and monitoring AI system behavior, addressing risks like shadow AI and data accuracy.

3. Machine Data Now Powering AI Models and Agentic Ops

Machine-generated data from logs, metrics, traces, and events became the fuel powering next-generation AI models in 2025. Organizations leveraged vast streams of operational telemetry to train specialized AI agents capable of understanding complex system behaviors and patterns. This shift enabled Agentic Ops. Agentic Ops allows AI agents to autonomously monitor, analyze, and act on infrastructure and application data in real time. By feeding machine data directly into AI models, teams achieved faster incident response, more accurate anomaly detection, and context-aware automation that could adapt to dynamic environments.

4. Massive Implications of AI on Technology Infrastructure

The explosion of AI workloads in 2025 forced organizations to fundamentally rethink their technology infrastructure. Traditional data center architectures struggled to support the computational demands of training large models and running inference at scale. This drove investment into specialized AI infrastructure, including high-performance GPU clusters, optimized networking fabrics, and distributed compute architectures. Organizations also prioritized infrastructure that could seamlessly support both AI training and inference workloads while maintaining flexibility for traditional enterprise applications. The result was a new generation of hybrid, AI-optimized infrastructure designed to handle the unique requirements of generative and agentic AI at enterprise scale.

5. Challenges and Innovation Around Sustainability

The rapid growth of AI raised significant concerns about data center energy consumption and carbon emissions. Innovations such as direct-to-chip cooling, liquid immersion, and strategic placement of data centers near renewable energy sources emerged to address sustainability concerns, aiming for fewer but more energy-efficient AI facilities.

6. Advances to Quantum Networking

Investment shifted toward quantum networking infrastructure, especially Quantum Key Distribution (QKD), as a practical method for enabling scalable quantum computing and security applications. Many efforts in 2025 focused on connecting multiple quantum processors to create distributed systems for applications.

7. Growing Partner Ecosystems and AI Talent Training

The complexity of AI deployment drove a surge in strategic partnerships and broad ecosystem collaboration. Organizations focused on reskilling internal talent to become AI specialists, emphasizing intellectual agility and cross-domain expertise over traditional specialized knowledge.

8. Popularizing MCP Servers and AI Infrastructure

Model Context Protocol (MCP) servers became popularized, standardizing how applications provide context to large language models, enabling richer AI interactions and integrations. Additionally, AI infrastructure that could cluster integrated networking and GPU compute resources gained traction, helping support agentic and generative AI workloads, as well as other enterprise-scale AI operations.

9. Even More AI Skills and Platforms To Choose From

The AI skills landscape broadened significantly, including advances in large language models (e.g., OpenAI GPT-5, Google Gemini 2, Mistral Large, etc.), multimodal AI, quantum machine learning, and AI safety and governance standards. Cloud platforms and inference services scaled to support faster, more reliable AI model deployment.

10. The Evolution of Ai-Driven Automation and Self-Healing Systems

AI agents and automation, as well as generative AI-powered human-in-the-loop workflows managed more complex IT and network operations in 2025. These AI-powered workflows included anomaly detection, alert management, and sometimes even automated remediation, leading to more reliable apps and services, self-healing systems, and resilient network environments.

What Can We Expect in 2026?

• A continued focus on federated AI analytics and AI-ready data fabrics to enable cross-domain insights.
• Agentic Ops will be moving more and more into production for enterprise companies throughout 2026, with some estimates around 1.3 billion active agents in 2028.
• Machine data will continue to take center stage, powering the next wave of AI innovation. A new kind of AI will be trained on both the language of humans and technological systems.
• The growth of unified AI assistants and agentic workflows supporting collaboration across IT, security, observability, and business teams.
• More focus on building trust, security, and governance around all AI ecosystems, supporting the rise of Agentic Ops. Data, visibility, security, and trust will be key as thousands of AI agents continue to join corporate networks.
• The expansion of AI-driven automation platforms with predictive assurance and proactive issue prevention.
• An increased emphasis on AI literacy, prompt engineering, and workforce transformation to adapt to AI-augmented workflows.
• Even more development of generative AI infrastructure clusters that can integrate networking and compute resources.

This perspective highlights how AI in 2025 evolved through efficiency, observability and security, sustainability, enhanced collaboration, and infrastructure innovation, setting the stage for continued transformative growth in 2026. To keep up with the latest AI trends in observability, security, and data management, please check out the Splunk AI page. Or, if you’re ready to try out one of our many AI solutions, sign up for a free trial or view an interactive product tour to check out these AI capabilities for yourself!