This was the most obvious indicator so far that artificial intelligence is not a novelty to consumers anymore, as it was not a new product or a scientific article, but a convention floor in Las Vegas. The undercurrent of everything Sundar Pichai, the chief executive of Alphabet, did take the stage at Google Cloud Next this week was that the company was making a determined, focused move into enterprise software and that that move was based on AI agents.
This did not come as a surprise to anybody who was keen enough. Yet the scale and particularity of the commitment, so laid out over three days, starting Wednesday, imparted to the technology a sort of institutional weight that it had hitherto lacked. Google is not merely providing AI tools to businesses on a trial basis. It is also making that market its main stage on which it will determine the future of its artificial intelligence business.
To the uninitiated, AI agents are software that can do much more than provide answers to questions or summarise documents. They are able to organize the course of actions, make decisions depending on the situation and perform activities independently, which in many cases are not guided by a human. They signify a major development of the chatbots that initially gained publicity, and they have become the main object of competition in the great AI laboratories. The opportunities are high. The complications are so, too, are the complications.
The merging of some of Google AI products into a new brand, Gemini Enterprise was one of the most impactful announcements at the conference. The heart of this rebrand is Vertex AI, a cloud platform that enables business customers to pick among an assortment of AI models and apply them to their own purposes. The rebranding is cosmetic to some extent, but the improvements that come along with it are not. Google has augmented the capabilities of Vertex AI to enable it to be a more powerful and comprehensive offering to companies that want to create and operate AI-powered workflows at scale. The point is easy to understand: in case you are a big company that struggles to understand how to make artificial intelligence work, Google would like to be the layer you are going to be based on.
The new line of governance and security features that were launched by the company are also tailored to AI agents. This is more than it may seem. The more able agents are, the more consequential they are. An agent capable of browsing the web, writing and executing code, making appointments or communicating with external services on behalf of a company is an agent as well, and can cause actual harm once it starts to act incorrectly, on bad information, or when it is controlled by a malicious actor. There is a real interest in the field among researchers, regulators, and even skeptical corporate technology buyers. As a way of dealing directly with those concerns, instead of making safety an add-on to its product, Google is attempting to inculcate governance and oversight as a core component of its product offering.
All this is timely. Conspicuous pivots toward enterprise customers have been made by OpenAI and Anthropic over the past few months, with the strategies of the two companies rebased around the organisations that constitute the most sustainable and scalable source of revenue. Brand recognition and data are useful in consumer products, but business contracts are what generate the type of predictable and large volume income that can keep a company operating at this level of investment. Google, having the relationships it has in cloud infrastructure and workplace software, is well-positioned to compete in that market, but it can not rest on its laurels. Its rivals have been changing at an abnormally rapid pace.
This story has a coding dimension as well. At the conference, one of the executives of Google stated that additional AI coding features will appear in May. The information is still scanty, yet this indicates another battle in the business arena. One of the most prized customers in the business software market are developers, and all of the big AI companies have spent millions of dollars developing tools aimed at speeding software development. GitHub Copilot is a product of both Microsoft and OpenAI that has already become an integral part of numerous engineering processes. Google is evidently determined to be able to provide something that will compete with that in a credible manner.
What is interesting about this scene, when seen through the small distance, is how fast the story about artificial intelligence has changed. A bit more than 2 years ago, the prevailing discourse was that chatbots, their amazing fluency, and their hallucinations at times, and their unpredictable usefulness. We are talking today of production infrastructure, enterprise contracts, governance structures and autonomous agents within corporate systems. The technology has not addressed all its issues. There are unanswered questions regarding reliability, oversight, and the true payback on investment. However, the industry has since outgrown the question of whether AI is useful or not and is now debating which AI version will be most useful to attract the most business.
