Ethan Mollick is a Wharton professor whose research and writing examine AI’s effects on work, education, and entrepreneurship. — Wharton Faculty Profile.

Infographic for "Lessons from Ethan Mollick".

On Co-Intelligence and Working with AI

  1. Invite AI to the Table: Try AI on a range of lawful, ethical tasks to learn its strengths and limits in your own work. — Big Think Interview.
  2. Keep a Human in the Loop: Check consequential AI output; plausible answers can still be wrong, and the user remains responsible for decisions. — Big Think Interview.
  3. Give AI a Role and Context: Describe the role, task, and context conversationally, while remembering that a useful human-like interface does not make the system a person. — Big Think Interview.
  4. Practice With Today’s Tools: Build experience with current systems because future versions may be more capable; Mollick presents improvement as an expectation, not a guarantee about every task. — Big Think Interview.
  5. Map the Jagged Frontier: In a BCG experiment, AI helped on tasks inside its capability frontier but made workers less likely to solve one task outside it. — Jagged Frontier Study.
  6. Use the Social Interface Carefully: Conversational instructions can help elicit useful work from AI, but Mollick warns against mistaking the interface for a human mind. — Big Think Interview.
  7. Treat AI as a Co-Intelligence: Mollick frames AI as a new general-purpose intellectual tool that can emulate thinking and writing while remaining fallible. — Co-Intelligence Excerpt.
  8. Beware Plausible Errors: Confident AI output can cause people to stop checking work; learn where the tool fails and retain active oversight. — Big Think Interview.
  9. Use AI for Idea Recombination: LLMs can connect unexpected concepts to generate practical ideas, though originality and usefulness still require evaluation. — Automating Creativity.
  10. Apply Domain Expertise: Experts can spot hallucinations, specify better tasks, and run cheap trials in their own work, making their knowledge valuable in AI use. — Latent Expertise.
  11. AI Gains Depend on the Task: In the BCG study, AI improved productivity and quality on the tested inside-frontier tasks but worsened correctness on one task outside its capability frontier. — Jagged Frontier Study.
  12. Look for Newly Possible Work: Ask which previously impractical task AI now lets a team attempt; Mollick presents this as a prompt for experimentation, not a universal outcome. — Behind the Tech Interview.
  13. Reassess Work That Changes in Value: Mollick asks organizations to notice tasks whose prior value may decline as AI capabilities improve. — Behind the Tech Interview.
  14. Verify Digital Media: Mollick’s inexpensive self-cloning experiment shows why apparently authentic audio or video deserves corroboration. — Cloning Yourself.
  15. Use Agents With Judgment: AI agents can plan and use tools to complete bounded tasks, but the value and appropriateness of those tasks still depend on human choices. — Real AI Agents.

On AI in Education

  1. Rethink Traditional Homework: Mollick argues that capable AI can solve many standard take-home tasks, so educators should test assignments and redesign assessment. — Homework Apocalypse.
  2. Do Not Rely on AI Detectors: Mollick warned that AI-writing detectors can miss edited outputs and falsely accuse students, making them a poor sole basis for discipline. — Homework Apocalypse.
  3. Explore Personalized Tutoring: Mollick sees promise in adaptive AI tutors, but a universal two-sigma improvement has not been demonstrated by the cited essay. — Future of Education.
  4. Keep Teachers Central: AI tutors can supplement instruction, while classrooms and teachers still provide practice, collaboration, social connection and support. — Future of Education.
  5. Generate Examples for Critique: Teachers can use AI to create many context-specific examples, then ask students to identify errors and compare applications. — How to Use AI to Teach.
  6. Explain, Then Check: AI can produce step-by-step explanations that students revise and improve; the exercise depends on evaluating its gaps and mistakes. — How to Use AI to Teach.
  7. Preserve Core Skills: Mollick expects a mix of AI-assisted and unaided assignments, much as calculators coexist with learning mathematics. — Future of Education.
  8. Make Students Critique AI Work: Assignments can require students to correct and improve AI-generated work, making their reasoning visible rather than outsourcing it. — How to Use AI to Teach.
  9. Set Mixed AI Rules: Mollick anticipates a practical mix of assignments requiring AI and assignments prohibiting it, while schools protect writing and critical-thinking skills. — Future of Education.
  10. Free Teachers for Interaction: If AI handles some explanation outside class, teachers may have more time for active learning and individual support; Mollick presents this as a potential design. — Future of Education.
  11. Require Transparent AI Use: In Mollick’s classes, students used AI and disclosed prompts while remaining accountable for errors; the aim was more ambitious work, not hidden cheating. — Wharton AI in Education Interview.
  12. Build AI Simulations: Mollick describes AI-powered educational simulations in which agents act as mentors, instructors and evaluators. — Wharton AI in Education Interview.
  13. Use Speech Tools as Teaching Inputs: Mollick reports using Whisper’s transcription of student speech and mixed accents in his teaching; its superiority to human hearing was his impression, not a controlled finding here. — Wharton AI in Education Interview.
  14. Treat VC Feedback as an Anecdote: Mollick reports that real VCs praised feedback from an AI-simulated investor in his class; this is an observation, not a comparative trial. — Wharton AI in Education Interview.
  15. Experiment Organizationally: Mollick viewed large gains in a specific BCG task experiment as a reason for organizations to test AI use promptly, not proof of gains in every workflow. — Wharton AI in Education Interview.

On Entrepreneurship and Innovation

  1. Entrepreneurship Can Be Learned: Mollick argues that founders can learn entrepreneurial skills rather than needing to fit a celebrity-founder personality. — Wharton Entrepreneurship Interview.
  2. Design Culture for Growth: A startup’s culture should support growth rather than become a barrier as the company scales. — The Unicorn’s Shadow.
  3. Test Startup Hypotheses: Mollick urges founders to show methodical testing, explain what experiments taught them, and adapt from the evidence. — Coller Venture Review Interview.
  4. Reject a Universal Founder Type: Mollick argues that no single personality reliably predicts founder success; some traits that encourage entry can hinder later listening and adaptation. — Survivorship Bias.
  5. Founders Are Not Necessarily Young: Mollick cites a mean founder age of 42 and warns that the celebrated twenty-something founder is not the typical successful model. — Survivorship Bias.
  6. Seek Diverse Inputs: An observational study found that employees with more diverse Twitter networks and stronger information-assimilation skills tended to submit better-rated ideas. — MIT Sloan Twitter Study.
  7. Pitch Professionals With Clarity: Mollick says professional investors respond more to preparation, organization and evidence than to showmanship alone. — Coller Venture Review Interview.
  8. Know When Showmanship Helps: Mollick distinguishes amateur investors, who may respond more to charisma, from professional investors who emphasize methodical evidence. — Coller Venture Review Interview.
  9. Listen to Users as Innovators: Mollick argues that people using a tool daily can cheaply discover valuable improvements that centralized teams may miss. — IT Revolution Q&A.
  10. Automate Tedious Tasks: Mollick suggests redesigning jobs around engaging work and assigning repetitive, low-value tasks to AI under human supervision. — In Praise of Boring AI.
  11. Take Boredom Seriously: Mollick cites research associating boredom with troubling behavior and proposes reducing needless tedium at work. — In Praise of Boring AI.
  12. AI Extends Intellectual Work: Mollick describes generative AI as a general-purpose tool that can augment or replace some human thinking and writing, unlike earlier physical tools. — Co-Intelligence Excerpt.
  13. Surface Hidden Employee Innovation: Mollick observes that workers may conceal useful AI experiments because of policy uncertainty, credit and job-security concerns. — Detecting the Secret Cyborgs.
  14. Align Incentives for Sharing: Mollick proposes rewards, trust and credible job-security commitments so employees will share valuable AI uses. — IT Revolution Q&A.

On the Future and Learning

  1. Expect an Adjustment Period: Mollick describes sleepless nights as he confronted AI’s implications for work and education; that was his experience, not a universal prerequisite. — Co-Intelligence Excerpt.
  2. Learn Through Repeated Use: Regular hands-on use helps reveal surprising AI strengths and weaknesses that a one-off query may miss. — Big Think Interview.
  3. Steer Toward Human Benefit: Mollick urges organizations to share positive AI uses that improve work and help people thrive, not just cost-cutting uses. — IT Revolution Q&A.

Learn more:

  1. Big Think: Ethan Mollick’s Four Guiding Principles for Leading with AI
  2. Penguin Random House: An Excerpt from Co-Intelligence
  3. Harvard Business School: Navigating the Jagged Technological Frontier
  4. Ethan Mollick, One Useful Thing: Automating Creativity
  5. Ethan Mollick, One Useful Thing: Latent Expertise—Everyone Is in R&D
  6. Microsoft Behind the Tech: Ethan Mollick on Future Opportunities with AI
  7. Ethan Mollick, One Useful Thing: A Quick and Sobering Guide to Cloning Yourself
  8. Ethan Mollick, One Useful Thing: Real AI Agents and Real Work
  9. Knowledge at Wharton: How Does AI Impact Education?—Interview Transcript
  10. Ethan Mollick, One Useful Thing: The Future of Education in a World of AI
  11. Ethan Mollick, One Useful Thing: The Homework Apocalypse
  12. Ethan Mollick, One Useful Thing: How to Use AI to Teach Some of the Hardest Skills
  13. Knowledge at Wharton: Forget the Myths—How Entrepreneurs Can Succeed in Real Life
  14. Wharton School Press: The Unicorn’s Shadow—Book Description
  15. Ethan Mollick, One Useful Thing: When Survivorship Bias Meets Superstitious Learning
  16. MIT Sloan Management Review: How Twitter Users Can Generate Better Ideas
  17. Coller Venture Review: Shifting the Entrepreneurial Paradigm to a Data-Driven View
  18. Ethan Mollick, One Useful Thing: In Praise of Boring AI
  19. Ethan Mollick, One Useful Thing: Detecting the Secret Cyborgs
  20. IT Revolution: Q&A with Dr. Ethan Mollick, Author of Co-Intelligence