The most common use of generative artificial intelligence is not drafting emails or writing code, but seeking emotional support and therapy, according to a June 2026 analysis published by Harvard Business Review. The category of 'therapy and emotional support' has doubled from 5% to 11% of total usage, surpassing technical troubleshooting and entertainment. This shift, observed from the United States to Indonesia and Italy, challenges the narrative of AI as primarily a productivity tool. In Jakarta, media report a growing habit of turning to chatbots rather than friends for personal problems, while Milan-based analysts note that AI is entering 'the most intimate zones of people's minds'.

Simultaneously, the labour market is recalibrating. In the United States, entry-level software development roles are vanishing, with AI cited for over 21,000 job cuts in April 2026 alone. Graduates from top American universities describe applying to thousands of positions without success, as employers demand years of experience for roles once open to newcomers. Bill Gates, in a series of interviews, has identified programmers, biologists, energy experts, and professional athletes as relatively insulated from automation, because they require human judgement, creativity, or emotional connection. In Indonesia, graphic designers, programmers, and customer service staff are being urged to transform their skills rather than face obsolescence.

Corporate experience is puncturing ambitious claims. Ford recalled 350 senior engineers after AI systems failed to meet design standards. Swedish fintech Klarna rehired human customer service staff after chatbots lacked empathy, and McDonald's suspended its AI voice-ordering trial following repeated errors. These reversals, reported in Italian and American business press, underscore that AI deployment often requires substantial human oversight and upfront investment, echoing the slow productivity gains seen during the electrification of factories a century ago.

The rise of AI as a confidant carries risks. A 2025 survey found nearly half of AI users with mental health conditions use chatbots for emotional support, and AI companion app usage grew roughly 700% from 2022 to mid-2025. Researchers warn of 'thinkslop'—a term for cognitive atrophy from outsourcing reflection to machines. A new technique called loop engineering, which instructs AI to maintain a continuous, relationship-focused dialogue, is being explored to make mental health chatbots more supportive, though it remains experimental. The next factual test of AI's economic impact will be the US Bureau of Labor Statistics' productivity report in August, which analysts in Washington and Frankfurt will scrutinise for signs that quality improvements from AI are finally being captured in official data.