Microsoft’s head of artificial intelligence, Mustafa Suleyman, has warned that a large share of white-collar work could be automated within the next 12 to 18 months as the company accelerates the development of advanced artificial intelligence systems aimed at the enterprise market.
Speaking in an interview with the Financial Times on Thursday, Suleyman said Microsoft is moving rapidly toward what he described as “professional-grade AGI,” referring to systems capable of handling a broad range of routine knowledge-work tasks with minimal human intervention.
Artificial general intelligence, or AGI, broadly refers to AI systems designed to match or exceed human performance across many intellectual domains, rather than being limited to a single task.
“White-collar jobs — essentially those sitting in front of computers, whether lawyers, accountants, project managers or marketers — most of these tasks will be fully automated by AI within the next 12 to 18 months,” Suleyman said.
From roles to workflows
Instead of focusing on replacing individual job titles, Suleyman said that Microsoft is developing AI agents, or software that can perform multiple workflows on its own. The AI agents are being developed to perform tasks such as contract writing, analyzing financial data, and managing complex administrative processes.
The strategy appears to indicate a shift from AI tools assisting employees to those that are able to function with a high degree of autonomy in all aspects of a business. Microsoft believes this will enable companies to use AI in their legal, finance, sales, and operations teams.
This warning has been given at a time when several major companies are undergoing a workforce restructuring process due to the increased use of automation. For instance, Salesforce has reportedly terminated around 1,000 employees this month in a move driven by AI, and Amazon, FedEx, and Ericsson have also announced job cuts partly due to efficiency and automation.
Push for in-house AI models
Looking ahead, Suleyman said the creation of customised AI models could soon become as simple as launching a blog or podcast is today. Within two to three years, he added, AI agents may be able to manage significant portions of institutional workflows with limited human oversight.
He also signalled a strategic shift toward greater internal control over AI development. Microsoft, he said, is working toward “true AI self-sufficiency,” with plans to expand its own model development and reduce dependence on external partners such as OpenAI.
According to Suleyman, Microsoft’s next generation of in-house AI models could begin rolling out globally as early as 2026, marking a major step in the company’s long-term AI strategy.
