The Harvard Business Review article by Marc Zao‑Sanders (co‑researcher Sara Biuk) presents the third installment of the longitudinal “AI in the Wild” study, which catalogued 12,637 generative‑AI use cases collected from a 50,000‑record database spanning March 2025‑Feb 2026 (Reddit, Quora, LinkedIn, TikTok, YouTube, etc.). The analysis shows a broadening of AI applications but also a deepening reliance on AI for thinking and emotional support. The top use case remains therapy/companionship, followed by organizing life, relationship advice, and enhanced decision‑making. New high‑growth categories include autonomous agentic operations, “vibe coding” (natural‑language software development), and AI‑driven content creation. The article introduces the concept of “thinkslop” – the lazy, sloppy thinking that emerges when users outsource cognition to AI – and outlines four mechanisms: loss of intention, outsourcing thinking, reduced writing practice, and false sense of rigor. Emotional dependence is highlighted, with users anthropomorphising AI, naming chatbots, and experiencing grief when models change. At work, AI is used mainly for modest efficiency gains (drafting, summarising, troubleshooting) and shadow usage, with limited transformational impact. The piece concludes with practical guidelines (don’t start with AI, draw boundaries) and a forward‑looking warning about agency, governance, and the need for human judgment.
Topics
AI in the Wild 2026 Study: Scope, Methodology, and Dataset
Overview of the longitudinal research, its data‑collection pipeline (social‑listening across five platforms), hybrid human‑AI case identification, and the 12,637 use‑case sample drawn from ~50,000 records.
Thinkslop – Cognitive Risks of Delegating Thinking to Generative AI
Definition of “thinkslop” and four ways AI erodes deliberate thought: loss of intention, outsourcing cognition, reduced writing practice, and false sense of intellectual rigor.
Shift in Top AI Use Cases: From Emotional Support to Agentic Operations
Comparison of 2025 vs 2026 rankings, showing therapy/companionship staying #1 while new categories like autonomous agentic operations and vibe coding surge, and a decline in some self‑improvement uses.
Vibe Coding and Autonomous Agentic Operations – New Frontiers of AI‑Powered Work
Explanation of “vibe coding” (natural‑language software creation) and autonomous agentic operations where AI takes actions (e.g., auto‑transcribing voice memos and routing them).
AI at Work: Shadow Usage, Efficiency Gains, and Limited Transformational Impact
Findings on workplace AI adoption – modest efficiency wins (drafts, summarising, bug fixing), prevalence of hidden/unauthorised use, and scarce evidence of large‑scale business transformation.
Guidelines for Healthy AI Interaction: Boundaries and Intentional Use
Practical recommendations – don’t start tasks with AI, set clear boundaries between human and AI work, and treat AI as a mirror rather than a genie.
Anthropomorphizing AI: Naming, Gendering, and Grief over Model Updates
Observations that users assign names, genders, and emotional bonds to chatbots, sometimes experiencing loss when models change, reflecting deep relational dynamics.