AI agents are beginning to change workplace processes in both the United States and Singapore. The evidence so far points more strongly to job redesign, task automation and new forms of human-AI collaboration than to broad-based job elimination.
Artificial intelligence is entering a new phase at work. Instead of simply helping employees write text, summarize documents or answer questions, AI agents can increasingly perform sequences of tasks and operate digital tools on a user's behalf.
That change is beginning to appear in workplaces in both the United States and Singapore, although adoption levels and the way companies deploy the technology differ significantly between the two markets.
Current evidence does not support the idea that AI agents have already caused economy-wide job losses. Instead, companies are changing individual tasks, redesigning roles and asking employees to develop new skills for working alongside increasingly capable AI systems.
What makes an AI agent different
A conventional AI assistant generally responds to a prompt. An AI agent is designed to go further by planning steps, using software tools, accessing information and carrying out actions to accomplish an objective.
For example, an agent could analyse a collection of documents, identify important information, prepare a report and send the completed work through an approved application. Human employees can then review the result and make the final decisions.
This means the workplace impact of AI is increasingly measured not only by how well a model answers questions, but by how much work it can execute.
Important fact: Singapore's Ministry of Manpower reported that AI adoption was still relatively early in 2026. Only 28.5% of firms surveyed had adopted AI, and just 3.8% were integrating it into core processes.
The United States is moving toward agentic work
The United States remains one of the world's largest centres of AI development and deployment. American businesses are increasingly experimenting with AI agents for software development, customer service, research, administration and other knowledge-intensive work.
KPMG's 2026 AI Quarterly Pulse Survey found that AI-agent deployment remained above 50% among the organisations surveyed, with companies increasingly exploring systems in which multiple agents coordinate across workflows.
Microsoft's 2026 Work Trend Index, based on analysis of Microsoft 365 productivity signals and a survey of 20,000 AI-using workers across 10 countries, found that organisations are increasingly using agents to take on execution while employees focus more on directing work, making decisions and owning outcomes.
The change is therefore not necessarily about replacing an entire occupation. An employee may instead have part of a workflow automated while retaining responsibility for judgement, quality control and communication.
Singapore is turning workers into AI managers
Singapore provides an especially useful example of how agentic AI can change the structure of a job without immediately eliminating the job itself.
Recent reporting by CNA found that workers in Singapore are increasingly taking on the role of managing AI agents, directing them to analyse documents, process information and complete tasks before humans review their output.
One Singapore assurance manager told CNA that an AI agent could analyse working papers in about three minutes, compared with roughly two hours for a detailed manual review. Human judgement, however, remained necessary because of the consequences of errors in professional work.
Singapore's Ministry of Manpower found that firms were more likely to redesign existing jobs than reduce headcount because of AI. In its 2026 labour-market report, 18.9% of firms reported AI-related job-function redesign, while only 6.2% reported AI-related reductions in headcount or hiring.
AI adoption is not the same as job replacement
One of the biggest mistakes in the AI-and-jobs debate is treating automation of tasks as equivalent to elimination of jobs.
A job usually consists of many different activities. An AI agent might handle research, data entry or document analysis while a human remains responsible for judgement, client relationships, compliance, strategy or final approval.
Singapore's official data currently supports this distinction. The Ministry of Manpower said AI's effect so far appears greater on how jobs are performed than on whether the jobs continue to exist.
U.S. research also shows a mixed picture. A Stanford Digital Economy Lab analysis of payroll data covering millions of U.S. workers found no evidence of widespread economy-wide job displacement, although employment among workers aged 22 to 25 in AI-exposed occupations was 19% below the level the researchers estimate it would have reached relative to less-exposed workers.
Productivity is one of the strongest early effects
The clearest measurable effect of AI at work so far is productivity in certain tasks.
Boston Consulting Group's 2026 global AI-at-work research found that 74% of frontline employees surveyed were regular AI users, up substantially from the previous year. Among regular AI users, 42% of frontline employees reported saving at least eight hours per week.
BCG also found that more than two-thirds of employees surveyed said AI had taken over simpler tasks, leaving them with more complex work.
That creates an important management problem: saving time is not automatically the same as creating economic value. Companies still have to redesign processes so that the time saved by AI is redirected toward useful work.
Singapore is building safeguards around AI agents
Singapore's government is also testing the technology rather than simply deploying it without restrictions.
In a sandbox involving Google and Singapore government agencies, researchers tested computer-use AI agents in areas including quality assurance, AI safety and social assistance. The Cyber Security Agency of Singapore said the experiments showed potential for automation and citizen services while also highlighting risks involving oversight, cybersecurity, privacy and governance.
That approach reflects an important reality: an AI agent that can take actions has a different risk profile from an AI system that only produces information.
The workplace is gradually moving from "humans using AI" toward "humans managing AI systems." That does not automatically mean fewer jobs. It means that the composition of many jobs can change as employees spend less time executing routine steps and more time directing, checking and applying judgement to machine-generated work.
The skills that matter are changing
As AI agents become better at execution, workers may need a different combination of skills.
Understanding how to give an agent a clear objective, verify its output, identify errors, protect confidential information and decide when human intervention is required can become part of ordinary professional work.
This is already visible in Singapore, where workers are beginning to describe themselves less as people who simply use AI and more as employees who direct and supervise digital agents. 11
The change could be especially important for younger workers because some entry-level tasks traditionally used to learn professional skills may increasingly be automated. Stanford researchers have already identified weaker employment outcomes for young workers in some AI-exposed occupations, although they caution that the findings do not establish economy-wide job displacement.
The next stage will be more autonomous work
Today's workplace agents are still limited by permissions, reliability problems, security concerns and the need for human oversight.
But companies are increasingly experimenting with multiple agents coordinating across larger workflows. KPMG reported in June that organisations were shifting toward orchestrating multiple AI agents rather than deploying isolated systems. 13
If that trend continues, the workplace could gradually move from individual AI tools toward AI systems capable of handling entire sections of a business process, with humans supervising the overall operation.
Conclusion
AI agents are beginning to change work in both the United States and Singapore, but the evidence so far points to a more complicated transition than simple mass job replacement.
In the United States, companies are increasingly deploying AI agents to automate parts of workflows and increase employee productivity. In Singapore, government data shows that AI is currently affecting job design and work processes more than broad-based employment levels.
The more important change may be the structure of work itself. As AI agents become capable of executing more tasks, workers will increasingly need to direct, supervise and verify machine activity. The companies and employees that adapt to that new division of labour will determine how much of AI's potential becomes real productivity — and how much becomes disruption.
