Agent confidence on the technical frontier

A survey of 300 global technology experts ranks 101 agent tasks, revealing where organizations are confident deploying agentic AI and where human oversight remains critical, according to a new MIT Technology Review Insights report sponsored by Microsoft [1]. The report identifies 2026 as an "inflection year" for aligning AI projects with strategic business objectives, citing Gartner [1]. IT infrastructure costs are projected to grow two to three times by 2030 even as budgets stay flat, per McKinsey [1]. Tech teams are already putting agents to work to automate tasks and manage workflows [1]. Confidence is highest for structured, measurable processes such as generating reports and boilerplate code [1]. Data workflows represent the breakthrough domain, with experts trusting agents most where structure provides a reliable foundation for decisions — including data quality monitoring, visualization anomaly detection, real-time data stream monitoring, and data profiling [1]. Human oversight remains a key factor of success [1]. Jeremy Winter, corporate vice president and chief product officer at Microsoft Azure Platform, stated: "As we design agents to operate within the same operational boundaries, identity systems, and governance models that teams already use, they start to behave more like the systems organizations already trust" [1]. The push toward agentic AI unfolds against a backdrop of rapid large language model advancement. LLMs, typically based on transformer architecture, are foundational to modern chatbots and can generate, summarize, translate, and analyze text [3]. Generative pre-trained transformers are pre-trained to predict the next word and then fine-tuned to follow instructions [3]. Microsoft Copilot, launched in 2023, utilizes a model built upon OpenAI's GPT large language models and was introduced as a built-in feature for Microsoft Bing and Microsoft Edge before expanding across Windows and Microsoft 365 [7]. Confidence in agents drops when tasks require deeper business context, which remains at an early stage of development [1]. The field of AI safety, which gained significant popularity in 2023, focuses on preventing accidents, misuse, or harmful consequences from AI systems and encompasses AI alignment to ensure systems behave as intended [6]. Researchers have expressed concern that AI safety measures are not keeping pace with the rapid development of AI capabilities [6]. Artificial general intelligence — a hypothetical AI that matches or surpasses human capabilities across virtually all cognitive tasks — remains a stated goal of companies including OpenAI, Google, xAI, and Meta [5]. A 2020 survey identified 72 active AGI research and development projects across 37 countries [5]. The experts surveyed for the MIT Technology Review report expect agent confidence to accelerate as experience deepens and business environments mature [1].