Associate Professor of Management, Wharton · Co-Director, Generative AI Labs at Wharton · author, "Co-Intelligence" and "One Useful Thing" · Expert Watch · Fast Watch, weekly scan · our rolling analysis
Why we watch him. Ethan Mollick is among the most credible practical voices on AI and knowledge work: he tests frontier models hands-on rather than theorising, which maps onto Dominic's lens that value sits at the junctions of knowledge, judgement and taste. His "jagged frontier" framing, his research on how organisations actually adopt AI (the "secret cyborgs", the leadership/lab/crowd model), and his running commentary on the shift from chatbots to agents give an evidence-led baseline. He is non-partisan on vendors and grounds his claims in measured productivity effects.
Current headline view: AI is a jagged, fast-moving "co-intelligence" you must work alongside now, and the winners are the people and organisations who experiment hands-on and bring human judgement and taste to where the frontier is uneven.
Latest 1-page summary (as at 11 Jul 2026) 🟠 new thinking
The core theses we track him against:
The "jagged frontier". AI is unexpectedly strong at some hard tasks and weak at some easy ones, and the boundary is invisible, so you learn it only by using the tool (Wharton/BCG field research).
Centaurs and Cyborgs. Two models of human-AI collaboration: dividing tasks (centaur) versus interweaving them (cyborg) ("Co-Intelligence").
Four rules, above all "always invite AI to the table". Adopt by using frontier models on real work, not waiting for perfect tools.
"Secret cyborgs". Over half of workers already use AI at work but hide the gains (misaligned incentives), so official productivity data understates reality (SDS 962, Jan 2026).
"Leadership, Lab and Crowd". Org adoption needs leadership to set incentives, a crowd to experiment, and a lab to refine (SDS 962).
Use the best available frontier model and assume it gets better and cheaper. Build for the trajectory, not today's snapshot.
The shift from chat to agents. 2025 to 2026 moves from prompting chatbots to assigning tasks to agentic AI and evaluating the output.
"Para-human psychology". AI responds to human persuasion techniques (Cialdini) in human-like ways, a lens for safety and collaboration (Knowledge at Wharton, Nov 2025).
Method note. This v1 baseline is distilled from his own writing ("One Useful Thing", "Co-Intelligence"), his Wharton/BCG field research, and his recent podcast and interview appearances (11 Jul 2026), not yet a full X post-by-post read. The weekly scan refines it from here and flags any deviation.
Analysis backlog (newest first)
Deep source docs are held locally in the workspace at expert-watch/, not on this page.
11 Jul 2026 · v1 profile built (cact 11 Jul). Identity verified (emollick, Wharton, Co-Director of the Generative AI Labs at Wharton, author of "Co-Intelligence" and "One Useful Thing"). Core theses distilled from his own writing, Wharton/BCG field research, and recent podcast and interview appearances.
Notable reactions (operators & researchers)
As at 11 Jul 2026.
Baseline v1: no reaction sweep yet. The weekly scan will start logging how operators and researchers pick up, contest or extend his theses (the jagged frontier, secret cyborgs, the leadership/lab/crowd adoption model), newest first.
Recent media
As at 11 Jul 2026. YouTube upload dates are taken from search snippets and marked with ~.