\n<\/p>\n
A year ago, engineers at Snowflake, the American cloud-based data platform, still spent part of their day on routine tasks \u2013 such as scanning dashboards to ensure systems were running smoothly and chasing colleagues for data to complete trend analyses.<\/p>\n
Now, says Qaiser Habib, the company\u2019s Toronto-based head of Canada engineering, AI agents handle much of that groundwork, allowing engineers to focus on higher-level decisions.<\/p>\n list of 4 items<\/span>end of list<\/span><\/section>\n Habib spends 20 to 30 hours a week interacting with five AI agents. Snowflake has built agents to review product design or to help on-call engineers to help during an outage or an incident, among other uses. He estimates the average engineer works with three or four agents daily, using them to carry out coding projects under human supervision.<\/p>\n \u201cYou don\u2019t have to bother a human for basic questions any more,\u201d Habib said, noting that he still collaborates with colleagues on more complex work, such as troubleshooting coding problems.<\/p>\n As companies experiment with AI agents \u2013 systems designed to plan, reason and carry out multistep tasks \u2013 the technology is beginning to reshape office hierarchies across the United States and Canada. Unlike chatbots, which respond to prompts, AI agents can adapt to changing contexts such as business goals and draw on reference tools including calendars, meeting transcripts and internal databases, to complete work with limited human oversight.<\/p>\n In some workplaces, AI systems are not just completing tasks but also assigning them to human workers. As the technology improves, AI agents are also beginning to manage each other. One agent might generate code, for example, while another reviews it for errors and fixes bugs before a human signs off on the final version.<\/p>\n These agent-to-agent workflows can help companies scale faster. But they also intensify concerns that AI is moving beyond assistance into supervision \u2013 and potentially, job replacement.<\/p>\n Anthropic recently expanded access to its cowork agents, allowing users without technical expertise to grant Claude \u2013 its AI assistant \u2013 permission to specific folders on their computers so it can read, edit, create and organise files autonomously.<\/p>\n The growing use of AI agents is transforming how organisations function around the world, even in companies that aren\u2019t focused on building technology products. For example, some companies are using AI tools to track performance, recommend promotions, role changes, and even identify roles for elimination.<\/p>\n The shift comes as white-collar jobs continue to disappear, particularly in the US. A slew of US employers have announced mass layoffs, mostly affecting entry-level and middle-management workers, and executives have pointed to automation and AI-driven efficiency as part of the rationale. When Amazon said in October that it planned to eliminate about 14,000 jobs, executives cited AI\u2019s potential to help the company operate with fewer layers and greater efficiency. UPS, Target and General Motors also announced deep cuts last\u00a0year, and this January saw more layoffs than any January in the US since 2009. Several more companies, including Pinterest and HP, continued to cite AI initiatives as part of the reason.<\/p>\n Goldman Sachs has estimated that 6 to 7 percent of US workers could lose their jobs due to AI adoption, with higher risks for computer programmers, accountants, auditors, legal and administrative assistants, and customer service representatives. Overall employment effects, the bank said in August, may be \u201crelatively temporary\u201d as new roles emerge.<\/p>\n Early predictions suggested AI would mainly replace entry-level technical jobs, and some experts tie recent high unemployment rates for new graduates to AI adoption. But the bigger disruption, said Roger Kirkness, founder of AI software firm Convictional in Toronto, is occurring in middle management.<\/p>\n His company\u2019s tools translate executive strategy into operational tasks \u2013 a role once handled by supervisors \u2013 delivering daily assignments and feedback to employees through a user-friendly inbox interface.<\/p>\n In companies of more than 50 people, \u201cwhere CEOs can\u2019t speak with each manager, our platform continually surfaces the context that the organisation has that is relevant to leadership decision-making\u201d, Kirkness told Al Jazeera.<\/p>\n This doesn\u2019t mean humans have become irrelevant. But there is growing pressure to reskill, and those who thrive in strategic thinking are better-positioned to adapt to AI-integrated work environments, Kirkness said.<\/p>\n \u201cPeople are basically becoming managers of their prior jobs,\u201d he said, because AI is now able to perform many of the tasks that previously fell within their roles. Instead of completing tasks such as coding or designing marketing assets, humans are focusing on higher-level strategy while monitoring AI systems, he added.<\/p>\n However, recent research indicates that job cuts reflect companies\u2019 anticipation of AI\u2019s potential, rather than its current ability to replace human workers fully.<\/p>\n A December Harvard Business Review survey of 1,006 global executives found that while AI has played little direct role in replacing workers so far, many companies have already cut jobs or slowed hiring in anticipation of its promised impact.<\/p>\n Most CEOs say they\u2019re still waiting on AI\u2019s payoff: 56 percent report no revenue or cost benefits so far, according to consulting firm PwC\u2019s latest Global CEO Survey of 4,454 executives across 95 countries and territories.<\/p>\n Stefano Puntoni, a behavioural scientist at the University of Pennsylvania\u2019s Wharton School, has found that AI usage is also already affecting workplace communication habits. His research shows employees are often more willing to delegate tasks to AI than to colleagues, which can help to reduce burnout. \u201cThere\u2019s no social cost,\u201d he said. \u201cYou don\u2019t worry about burdening an AI.\u201d<\/p>\n Still, Puntoni argues the biggest barrier to adoption is psychological, not technical. Even effective systems can fail if workers do not trust them. Generative AI, he said, can threaten employees\u2019 sense of competence, autonomy and connection.<\/p>\n \u201cIf workers feel threatened, they may want the system to fail,\u201d Puntoni said. \u201cAt scale, that guarantees failure.\u201d<\/p>\n In other words, deploying AI primarily as a cost-cutting tool can backfire. Layoffs framed as efficiency gains may reduce cooperation and limit the productivity benefits companies hope to unlock with technology, Puntoni said.<\/p>\n Trust, Kirkness agreed, is the real constraint. To build staff confidence in the tools it sells \u2013 and to avoid layoffs \u2013 Convictional adopted a four-day workweek, framing it as a way to share AI-driven productivity gains with employees.<\/p>\n \u201cMass layoffs in the name of automation destroy trust,\u201d he said.<\/p>\nRecommended Stories <\/h2>\n
The leaner office<\/h2>\n
Middle management squeezed<\/h2>\n
Trust and control<\/h2>\n
The human premium<\/h2>\n