The world’s largest technology companies are pouring money into a new kind of software, AI “agents” that can operate a computer on their own to get tasks done. At its Build developer conference in early June, Microsoft Chief Executive Satya Nadella said the era of operating systems and individual apps is giving way to “agent-first” computing, in which artificial intelligence acts across programs and devices rather than waiting for a person to click through each step. It was one of the clearest signals yet that the industry sees self-directed software as the next major shift in how people use computers.
For years, tech companies tried to build digital helpers for routine online chores, things like assembling a shopping cart, tracking an order, or planning a trip. Those early assistants could answer questions or draft a little text, but they could not really do much. The new generation is different. These agents can take action on their own: browsing the web, filling out forms, clicking buttons, and writing and running computer code to complete multi-step jobs with limited human help.
The capabilities have advanced quickly. OpenAI’s Operator now succeeds on roughly 87% of complex web-browsing tasks, according to company testing. Anthropic’s Claude can control a computer directly, write software on its own, and coordinate teams of smaller “sub-agents” working in parallel. Google’s Project Mariner can juggle about ten tasks at once on cloud-based machines. What was a research demonstration a year ago is now being sold as a product.
The competition has settled into a five-way race among OpenAI, Anthropic, Google, Microsoft, and Amazon, each spending heavily and each holding a different advantage. At Build, Microsoft released free open-source tools for developers to build agents, added an agent called Scout to its Copilot assistant, and said Copilot would now route each task to whichever model fits best, including those from OpenAI, Anthropic, and open-source providers. Google, meanwhile, has rebranded much of its enterprise AI strategy around agents and introduced multi-agent tools that can, for example, have one agent build a website while another creates the brand artwork.
The clearest battleground is software coding, which has become the leading business use for agents because the agents themselves are built out of code. Anthropic has surged ahead in this area on the strength of Claude Code, prompting OpenAI to shift much of its focus from consumers to businesses with a rival product called Codex. Enterprise sales now account for a significant share of OpenAI’s revenue, and its coding tools have attracted millions of users. Google is using its scale to compete aggressively on pricing even as Chief Executive Sundar Pichai recently acknowledged that the company remains behind some rivals in parts of the agent race.
Adoption inside companies is moving fast, at least according to industry surveys. Research cited across the sector suggests a large majority of business teams are already experimenting with agents, with many organizations deploying a dozen or more. Helping that spread is a shared technical standard known as Model Context Protocol (MCP), originally developed by Anthropic, which allows agents from different companies to connect to the same applications and data sources much like devices plugging into a common electrical outlet.
The reason the technology giants are betting so heavily is simple. If agents can reliably handle office work, customer service, scheduling, research, software development, and routine digital tasks, they could unlock enormous productivity gains and reshape the software industry. Whoever controls the agent people and businesses rely upon by default could ultimately control the most valuable layer of computing, much as smartphone makers once controlled the app economy.
The vision is a shift from people using apps to people simply telling an agent what they want done.
There are real reasons for caution. The technology remains early and far from perfect. Agents still make mistakes, can take incorrect actions without supervision, and require oversight for important tasks. Despite the enthusiasm, the financial results remain mixed. Industry research shows that while a large share of chief executives rank artificial intelligence among their highest priorities, only a relatively small percentage report substantial financial gains so far. Reliability and trust remain the biggest obstacles separating impressive demonstrations from everyday business use.
For now, the AI agent race represents the technology industry’s most expensive wager since the arrival of the smartphone. Microsoft, Google, Amazon, OpenAI, and Anthropic are betting that within a few years software capable of acting independently will become the normal way people get work done. Whether that prediction transforms daily life or proves overhyped will depend on something less exciting than the demonstrations on stage: whether these agents can consistently do the job correctly, day after day, without someone constantly watching over them.
JBizNews Desk — Technology
© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.