.png)
2025 was the year the "chatbot" died and the "agent" was born. What started in January with a disruptive punch from Deepseek ended in December with a unified industry protocol, signaling a permanent shift in how we work. 2025 saw AI transition from a novelty writing tool into a proactive, enterprise-grade teammate. Here is how the last 12 months redefined the possible in AI.
In January, the world was rocked by the announcement of Deepseek. Yeah, that was still this year. After China demonstrated that it didn’t need the latest and greatest in GPUs to throw heavyweight punches, western AI companies woke up, with OpenAI launching its “Operator”: a harbinger of the agentic age.
In February, this vision started to become clear, with OpenAI’s Deep Research tool, which finally enabled marketers to actually use AI for tasks beyond writing social media copy. Anthropic also kept pace, releasing Claude 3.7, which quickly became a staple of marketers’ toolkits for its remarkably human-like tone.
March was the “Ghibli” month, with GPT-4o’s update showcasing the immense potential of AI for image generation. This model was one of the first to truly create “pixel-perfect” images, a trend that would continue to evolve dramatically. This was also a month of upset for the SEO old guard, as AI Overview began rolling out, slowly but surely starting to pick away at organic search numbers, even for top-tier websites.
As the quarter rolled over, we saw further consolidation. From xAI’s merger with X, to GPT 4.5 passing the Turing test, to the rise of AI assistants in smartphones and personal computers (like Microsoft’s Copilot push), the AI industry was prioritizing rapid progress above all, perhaps at the expense of practical usability.
In May, we saw deeper integration of AI into browsers, AI agents unsuccessfully running a company, AI Search Overview becoming even more prominent, the launch of true AI-powered video editing (watch this space), and the democratization of AI agent creation.
June brought us the much-maligned “launch” of Apple Intelligence, which didn’t actually go live for months, and the rise of Content Credentials, a crucial step toward better authenticating AI-generated content. In July, we witnessed the first “true” mainstream AI agent in the form of OpenAI’s ChatGPT agent (though we at Yarnit beat them to it), and a surprising acquisition of Superhuman by Grammarly, demonstrating that simply being a first mover wouldn’t guarantee dominance in the AI wave. AI infrastructure was also quietly being fortified, thanks in no small part to Trump’s executive push to surpass China in the AI race.
August and September saw even more shakeups and an intense focus on “Enterprise General Intelligence”. Alongside shockers like Anthropic taking over OpenAI in market share for the enterprise sector, and Meta’s 4th team restructuring in as many months, studies surfaced showing the tangible impact of AI on the bottom line, setting gears turning in decision-makers' heads.
October was arguably the month that cemented 2025 as the year of “agentic AI”. Apart from Hubspot’s AI agents, Salesforce took the limelight with Agentforce, which finally brought agents to the enterprise in a way that seamlessly plugged into existing workflows. OpenAI also launched Atlas, throwing its hat into the AI browser ring, and Microsoft immediately responded with updates to Edge. Netflix went all in on generative AI, a perfectly timed move considering the general access launch of Google’s Veo 3 video algorithm.
November marked the true rise of AI search, with Gemini’s integration into Google Search, and one of the biggest acquisitions of the year when Adobe bought Semrush for $1.9 billion. Agent 365 saw agents arriving on personal computers, and Nano Banana Pro set the new gold standard for AI-generated images.
And finally in December, Google rolled out updates to AI Mode, Disney invested $1 billion into OpenAI as their first content licensing partner, and the essential Model Context Protocol was finally adopted by Anthropic, OpenAI, Google, Microsoft, and more.
This collective move not only closed out a year of relentless innovation but also set a unified, powerful context for the next year, setting the stage for the dawn of truly ubiquitous and standardized agentic AI.
