Every year in sunny California, hundreds if not thousands of the brightest minds in tech gather for a conference where the latest round of announcements are given and deep-reaching predictions are made for the future of the space. Ten years ago, Google chief executive Sundar Pichai famously took the stage to announce the company was now 'AI first,' something which in the murky past of 2016 seemed a far-flung statement at best. Now, well into the AI revolution, Mr Pichai has laid out his clearest vision for an 'agentic' AI future yet, describing a world where artificial intelligence systems won't just answer questions but will increasingly act on people's behalf.

Speaking at Google's annual I/O event on Tuesday and Wednesday, where news.com.au was a guest, Mr Pichai repeatedly returned to one central theme: the next era of AI will be defined by autonomous digital agents capable of carrying out complex, long-running tasks for users. 'I think this year at Google I/O laid the foundation for agentic transformation across the products,' Mr Pichai said during a Q&A session alongside Vice President Liz Reid and Chief technology officer Koray Kavukcuoglu. 'At some point, you would query, get some information back. From there, you move onto having ongoing dialogue, conversation with these products to these products being able to actually take meaningful actions and make your lives easier on that basis.'

Agentic AI refers to systems capable of independently carrying out tasks over time, often across multiple apps or services, with limited human supervision. Rather than simply responding to prompts, these agents can plan, reason and execute actions in the background. At I/O, Google unveiled several products designed around that concept, including Gemini Spark which was described by the company as a '24/7' AI agent, alongside 'information agents' for Search and upgrades to its Antigravity AI development platform.

In an interview with The New York Times following the conference, Mr Pichai made it clear Google sees these developments as part of a much bigger trajectory toward artificial general intelligence (AGI) – AI systems with capabilities comparable to or exceeding human reasoning. 'There is inevitable progress toward A.G.I. that's happening,' Mr Pichai told the Times. 'I absolutely am sure that the technology is making foundational progress toward A.G.I.' While he stopped short of predicting exactly when AGI might arrive, he suggested recent advances had accelerated expectations. 'The rate of progress over the last one to two years has made me feel it's on the closer side than not,' he said.

That future is increasingly being shaped around AI agents. But what does an agentic future look like? Lots of data centres for a start – but we knew that. Agentic AIs are even more token hungry, and vast infrastructure is needing to be implemented in order to fuel it. The short of it coming back to users is because these tools are becoming smarter and more efficient – being able to work around the clock and create subprograms capable of executing tasks – the problems they are able to solve likewise become more numerous and complex. Before Google search, someone maybe consulted an encyclopaedia (or library even!) and it has since become a kind of 'one stop shop' for information. The continuation of that progression means Google and other AIs are increasingly able to solve users' problems themselves rather than middle-manning to other parties.

Google's push also comes amid intensifying competition with OpenAI, Anthropic and Chinese AI companies, particularly in coding and autonomous agents. Mr Pichai acknowledged Google had fallen behind rivals in some areas. 'When it comes to agentic coding with tool use, instruction following and long-horizon tasks, I think we are a bit behind at this moment,' he told the Times. 'But we are hard at work.' Still, he insisted Google's scale and suite of platforms, including Search, Android, Chrome, YouTube and Cloud, gives it a unique advantage in bringing AI to mainstream users. 'We are the only large company that is actually at that frontier,' he said.

One of the strongest themes running through both the conference and Mr Pichai's interviews was the idea that AI will fundamentally alter how people interact with technology itself. Google Search, for example, is evolving away from the list of links users have known for decades and towards a conversational assistant capable of tracking tasks, monitoring information and proactively surfacing updates. 'In an A.I. Mode, in an agentic mode, these things are going to do a lot more for you than what we were able to do for users 10 years ago,' Mr Pichai said.

But the Google chief also acknowledged growing public anxiety around AI, particularly concerns about jobs and the speed of technological change. 'A.I. is viewed as the most profound technology humanity will ever work on,' he told the Times. 'People, rightfully so, are anxious about the future that this technology will bring.' Even so, Mr Pichai argued the benefits would ultimately outweigh the disruption, comparing AI's impact to previous transformative technologies such as spreadsheets and smartphones. 'People are going to be more productive, they will have more time for leisure,' he said. 'There are many positive dimensions to it that are maybe not being talked about.'

The concept of agentic AI is not entirely new; it builds on earlier ideas of autonomous agents and robotic process automation. However, the scale at which Google is now pursuing it marks a significant shift. With the introduction of Gemini Spark, Google aims to offer a persistent personal assistant that can handle tasks like scheduling, email drafting, and even complex research across multiple data sources. This is a leap from the current generative AI chatbots that require continuous user prompting. The 'information agents' for Search represent another evolution: instead of returning a list of web pages, these agents will proactively monitor topics chosen by the user and deliver summaries, alerts, and recommended actions.

Underpinning all of this is Google's Antigravity AI development platform, which received upgrades to facilitate building and deploying agentic systems. Developers can now create agents that use multiple models, call external APIs, and maintain state across sessions. This combination of consumer and enterprise offerings positions Google to compete aggressively in the AI arms race. Yet, analysts note that OpenAI's ChatGPT with plugins and Anthropic's Claude with tool use have already demonstrated agentic capabilities. In the coding domain, companies like GitHub Copilot and Replit have set high bars for autonomous code generation and execution.

Pichai's acknowledgment of being behind in agentic coding is notable. It suggests that Google may have invested heavily in large language models but is now playing catch-up in the practical application of those models to real-world tasks. However, Google's advantage lies in its vast ecosystem: Search data, YouTube content, Android user base, and Google Cloud infrastructure. By integrating agentic AI into these platforms, Google can offer seamless experiences that third-party agents cannot replicate. For instance, an agent built into Android could directly access user location, calendar, and contacts to make restaurant reservations, while an agent in Chrome could manage passwords and autofill forms intelligently.

The broader implications of agentic AI extend to the economy and society. If these agents become as ubiquitous as smartphones, they could reshape industries such as customer service, personal assistance, education, and healthcare. However, they also raise questions about privacy, security, and accountability. Who is responsible when an agent makes a mistake? How will users control what agents do on their behalf? Google has not yet provided detailed answers, but the company's history suggests it will prioritize user control and transparency, at least in its public communications.

As the AI landscape rapidly evolves, the race toward AGI is central. Pichai's comments suggest that Google believes agentic AI is a stepping stone to AGI. By building systems that can handle longer tasks and reason more deeply, Google is laying the groundwork for machines that can learn and adapt across domains. Whether that goal is achievable within a decade or more remains debated, but the momentum is undeniable. With every major tech company pushing aggressively, the next few years will likely see dramatic advances in what AI can do autonomously.

Source: news.com.au News