what AI tool to pick when

Reframes the common question "Which AI should I use?" into a more productive one: "What am I asking AI to do?" Introduces a two-axis framework — verification feasibility (easy to hard) crossed with solution space (bounded to divergent) — that maps research tasks onto three zones of engagement: Automate (bounded tasks you can easily verify, like exploring datasets), Augment (middle-ground work like literature reviews), and Collaborate (divergent tasks that resist easy verification, like making sense of complexity).

For each zone, walks through three layers of engagement: what the task actually is, how to set up the work, and what model to use — distinguishing between chat platforms, AI-enhanced tools, and AI agents. Closes with two takeaways: the frontier keeps moving (reasoning, math, and research capabilities expand fast), and your choices shape what you see (these are analytical decisions, not just productivity decisions).

Part of the Learning to Research with AI series organized by the IDEA Lab at Imperial Business School.