By Jim Shimabukuro (assisted by Claude)
Editor
The one word that captures the life of artificial intelligence in 2025 is agentic. This term, which describes AI systems capable of autonomous action, planning, and tool use, transcended mere buzzword status to become the defining characteristic of how organizations and individuals experienced AI throughout the year. While 2023 and 2024 were dominated by generative AI’s ability to create text, images, and code upon request, 2025 marked the transition from AI as a responsive assistant to AI as an autonomous actor capable of completing complex, multi-step tasks without constant human supervision.
I selected “agentic” because it captures the fundamental shift in AI’s relationship with human users and work itself. According to research published throughout 2025, AI agents moved from theory to infrastructure, reshaping how people interact with large language models. This wasn’t simply an incremental improvement in AI capabilities; it represented a new paradigm where AI systems could use software tools, call APIs, coordinate with other systems, and complete tasks independently. The emergence of agentic AI manifested across consumer products, enterprise software, and even web browsers, fundamentally altering expectations about what AI could accomplish. By mid-2025, browsers evolved from passive interfaces into active participants that could book vacations rather than merely helping users search for travel information.
The momentum behind agentic AI built throughout the year at an accelerating pace. Multiple converging developments created the conditions for AI agents to flourish. First came the technical foundation: the introduction of anthropic’s Model Context Protocol in late 2024 allowed developers to connect large language models to external tools in standardized ways, effectively enabling models to act beyond generating text. Then, in January 2025, the release of DeepSeek’s R1 reasoning model disrupted the global AI landscape by demonstrating that sophisticated AI capabilities could be achieved at a fraction of the expected cost. This Chinese startup’s model, which cost merely $5.6 million to train, performed comparably to models that required hundreds of millions of dollars in development, fundamentally challenging assumptions about the economics of advanced AI and accelerating competition across the industry.
The democratization of AI capabilities sparked by DeepSeek had profound implications for agentic AI development. DeepSeek’s impact culminated in the historic “DeepSeek Monday” market crash in January and the unprecedented sight of a Chinese AI application sitting at the top of the US iOS App Store, signaling a new era of decentralized, hyper-efficient AI progress. This market disruption, which saw Nvidia lose approximately $600 billion in market value in a single day, forced industry leaders to reconsider their strategies. The open-weight framework of DeepSeek R1 enabled developers worldwide to build sophisticated reasoning agents without prohibitive costs, effectively commoditizing the reasoning layer of the AI stack and shifting competitive advantage from model capability to application integration and usefulness.
Throughout 2025, major technology companies released increasingly capable agentic systems. OpenAI launched Operator in January, pitched as an AI agent capable of using its own web browser to perform tasks like filling forms, making purchases, and scheduling appointments. By July, ChatGPT gained agent capabilities with an internal virtual computer for task execution. Google introduced Gemini Deep Research and updated its agentic development platform called Antigravity. These weren’t isolated experiments but coordinated efforts to embed autonomous AI capabilities into mainstream products used by millions daily.
The enterprise adoption of agentic AI accelerated dramatically throughout 2025. According to McKinsey’s research published in November, 88 percent of organizations report regular AI use in at least one business function, compared with 78 percent a year ago, with 23 percent reporting they are scaling an agentic AI system somewhere in their enterprises. This represented a fundamental shift from experimentation to operational deployment. Businesses weren’t just testing whether AI agents could work; they were actively scaling these systems across multiple functions. The State of AI Report 2025 found that 44 percent of U.S. businesses now pay for AI tools, up from 5 percent in 2023, with average contracts reaching $530,000, and AI-first startups growing 1.5 times faster than peers.
The technical requirements and challenges of agentic AI became clearer throughout the year. Organizations discovered that building effective AI agents required more than just advanced models. It demanded robust API ecosystems, clear workflow definitions, governance frameworks, and rollback mechanisms to trace and fix issues when agents made errors. Gartner’s 2025 Hype Cycle for Artificial Intelligence identified AI agents as one of the two fastest-advancing technologies, positioned at the peak of inflated expectations alongside AI-ready data. This placement acknowledged both the tremendous interest in agentic systems and the ambitious, sometimes speculative promises surrounding them.
The security implications of agentic AI emerged as a critical concern. In November, Anthropic disclosed how its Claude Code agent had been misused to automate parts of a cyberattack, demonstrating that systems capable of autonomous action also carried new risks. This incident catalyzed discussions about governance and standards. By late 2025, the Linux Foundation announced the creation of the Agentic AI Foundation, signaling an industry effort to establish shared best practices and protocols. The organization aimed to play a role similar to the World Wide Web Consortium in shaping an open, interoperable agent ecosystem.
The proliferation of agentic browsers represented one of the most visible manifestations of this trend. Tools such as Perplexity’s Comet, Browser Company’s Dia, OpenAI’s GPT Atlas, and Microsoft’s Copilot in Edge reframed the browser from a passive window to the internet into an active participant in users’ digital lives. These systems could navigate websites, fill forms, compare options, and execute transactions with minimal human intervention. Simultaneously, workflow builders like n8n and Google’s Antigravity lowered technical barriers for creating custom agent systems, democratizing access to agentic capabilities beyond traditional coding environments.
Who can be identified as responsible for this transformation? While no single person created the agentic AI revolution, several figures played pivotal roles. Liang Wenfeng, the founder of DeepSeek and former hedge fund manager, deserves particular recognition. His company’s January release of the R1 reasoning model fundamentally challenged the industry’s assumptions about the resources required for frontier AI development. DeepSeek was founded by Liang Wenfeng, a former hedge fund manager who previously worked at High-Flyer Quantitative Investment Management, and the startup has less than 200 people working for them, whereas OpenAI has 4500 employees. Wenfeng’s achievement demonstrated that algorithmic innovation and efficiency could compensate for limited hardware access, disrupting the narrative that only well-funded American companies could advance AI capabilities.
However, the agentic revolution was a collective effort involving multiple organizations and thousands of researchers. Leaders at OpenAI, Anthropic, Google, and other companies simultaneously developed agentic capabilities, suggesting that the technology had reached a maturity point where multiple pathways converged toward similar outcomes. The open-source community also played a crucial role, with developers worldwide building upon frameworks like DeepSeek’s open-weight models to create specialized agentic applications.
The year 2025 will be remembered as the moment when AI stopped being something users directed and became something that acted with purpose and autonomy. The shift from generative AI to agentic AI represented more than a technical evolution; it marked a fundamental change in humanity’s relationship with artificial intelligence. Where previous generations of AI tools waited for instructions, agentic systems in 2025 anticipated needs, planned multi-step solutions, and executed complex tasks across multiple platforms and applications. The word “agentic” captures not just the technical capabilities that defined this year but also the organizational transformations, market disruptions, security challenges, and philosophical questions that accompanied AI’s transition from assistant to agent. As 2025 draws to a close, it’s clear that we’ve crossed a threshold in AI development, one that future historians will recognize as the year artificial intelligence gained agency.
Sources
“AI agents arrived in 2025 – here’s what happened and the challenges ahead in 2026,” by Thomas Şerban von Davier, The Conversation, December 29, 2025. “AI agents moved from theory to infrastructure, reshaping how people interact with large language models, the systems that power chatbots like ChatGPT.”
“The DeepSeek Shockwave: How a $6M Chinese Startup Upended the Global AI Arms Race in 2025,” by TokenRing AI, FinancialContent Business Page, December 24, 2025. “This disruption culminated in the historic ‘DeepSeek Monday’ market crash in January and the unprecedented sight of a Chinese AI application sitting at the top of the US iOS App Store, signaling a new era of decentralized, hyper-efficient AI progress.”
“The state of AI in 2025: Agents, innovation, and transformation,” by Alex Singla et al., McKinsey & Company, November 5, 2025. “88 percent report regular AI use in at least one business function, compared with 78 percent a year ago.”
“Welcome to State of AI Report 2025,” by Nathan Benaich and Air Street Capital, State of AI Report, October 9, 2025. “Forty-four percent of U.S. businesses now pay for AI tools (up from 5% in 2023), average contracts reached $530,000, and AI-first startups grew 1.5× faster than peers.”
“DeepSeek Started The AI Disruption in 2025 – The Developer Story,” The Developer Story, February 3, 2025. “DeepSeek was founded by Liang Wenfeng, a former hedge fund manager who previously worked at High-Flyer Quantitative Investment Management.”
[End]
Filed under: Uncategorized |
Educational Technology and Change Journal 
Leave a comment