Ethan Mollick
Related people
Episodes
- This AI Expert Argues We're Underestimating AI's Productivity Impact — With Ethan Mollick
- 693. What’s Our Job as Humans Anymore?
Notes
Big Technology Podcast · This AI Expert Argues We're Underestimating AI's Productivity…
-
Will Personal AI Agents Flood Customer Service?
Ethan Mollick says a perfectly legal distributed denial of service on customer service departments is coming, authorized by millions of consumers, once it is widely known that Muse, Dot, and similar agents can make unlimited requests for old refunds and negotiate deals. Banking and customer service depend on friction: it is hard to make a phone call, annoying to file something, and it takes effort to gather paperwork.
Proactive agents already do the work at scale. His Dot cancelled several long-term subscriptions he could not get out of, and he does not know how many systems it spun up. As the tools become widely available, companies are not ready. Apollo Research pointed out a related thre
-
Will AI Replace All Human Labor by 2035?
Ethan Mollick says nobody really knows what happens to jobs. The setting is like an industrial revolution: some job categories will be invalidated in unexpected ways, some will transform, and some will grow. AI can program. Demand is up at least for senior developers because more coding is possible. A small organization can have one coder instead of a team of 20. Programmer jobs may grow, and the work is also shifting from writing code to managing code writing.
The AI labs are convinced they will replace all human labor by 2035. It is worth thinking hard about the safety nets to build during the transition. Even when the Industrial Revolution worked out well, it was grim to live through for
-
How Much of America's 20th-century Edge Came From Management?
Ethan Mollick says 40 percent of the advantage of American companies in the 20th century came from better management, according to a paper by Raffaella Sadun and other Harvard researchers. Companies used to experiment with how to organize themselves and get work done. American manufacturing and firm organization transformed. That exploration has not happened as much in the last 30 years.
Firms now let SaaS vendors dictate how to run part of the business, hire consultants, and point R&D at customers. A lot of organizations have very little muscle memory left for organizational change. It is rare to see a company fundamentally redivide how work operates, how roles operate, and what mandate pe
-
Are Large Companies Seeing a Return From AI?
Ethan Mollick says surveys show 85 percent of companies had positive ROI from AI as of last year. The major transformation has not happened yet, except it is starting in some organizations. A company with 100,000 employees still has to map processes, decide who gets tokens, and adopt agentic models that are less than a year old.
Productivity and cost savings are already there. Colleagues at Wharton surveyed a thousand large companies, and as of last November, 85 percent of large companies with over a billion dollars in revenue reported positive ROI from AI use. Detailed breakdowns of cost savings versus spend are not in that survey.
-
Why Don't Companies See AI Productivity Gains Yet?
Ethan Mollick says you do not get organizational productivity gains instantly from individual ones. A person can produce 100 times more code than before, which is great for a solo entrepreneur. Inside a company running two-week sprints, coordinating with marketing, and holding daily stand-ups, that 100 times gain becomes a barrier. Nobody has time to deal with all of that code.
Shipping stats show the same gap: 20 times more code produced, and 30 percent more products shipped. Organizations have not adapted to AI yet. Alex Kantrowitz cites App Store figures of more apps and fewer reviews. A flood of novels, games, and apps is one pattern. Studies of products shipped show a huge increase in
-
Why Are AI Token Prices Falling?
Ethan Mollick says AI token prices face real downward pressure because a huge move for token efficiency is underway and costs will decrease even as more compute comes online. Lower cost means AI does more stuff for less money in a very short period of time, so long-term token costs do not worry him.
A chart from Andreessen Horowitz compares the price of compute with the price of tokens. Compute has stayed steady and risen a little, while the price of tokens is falling dramatically. Alex Kantrowitz calls the drop the most deflationary technology he can remember in the price paid for the technology. Use is still rising because agent-style work requires more tokens, and a Fable class model can
-
Do AI Tutors Actually Help Students Learn?
Ethan Mollick says a universal tutor of high quality now exists in AI, and when the tool is set up as a tutor the evidence seems to show positive effects in a meta-analysis and in individual RCTs at Stanford and in Kenya. One-on-one tutoring adapted to the student helps people learn. The tool does not tutor by default, because it wants to do the homework.
People remember having to take Khan Academy videos outside class in a flipped classroom. The flipped classroom is suddenly effective with AI tutors teaching outside class. Inside class, active learning takes over through case discussions, exercises, debates, and discussions, usually without AI in the room, though sometimes AI is used as a
-
Why Do Superforecasters Keep Getting AI Wrong?
Ethan Mollick says superforecasters keep getting AI wrong because an exponential is hard to predict. The labs have been right. Development has stayed on an exponential curve with a fast doubling time. He thought slowdowns were a real possibility a few years ago and no longer thinks they are.
The market has grown faster than the labs expected. The technology has surprised him, while the lab leaders are on target for where they think they are going. He is not sure they were surprised by the barriers they have overcome. Reasoning is something that they were not expecting to work as well as it has, and reasoning has been core to the math breakthroughs. People expected AI to be as good at math a
-
When Did Practical AI Agents Become Possible?
Ethan Mollick says practical agents were not possible until models reached Opus 4.6 level or 4.5 level around November and December of last year. Agents became practical at that point. Co-Intelligence, his last book, was about the chatbot era because that was as good as models were. The newer book is about the agentic era because that is where models are today.
Product formats are shifting for three reasons at once. The study of innovation calls the first an era of ferment: nobody knows the right form, so ideas churn and get copied. OpenClaw came out in January as personal agents, and the form took root. Muse and Dot are claw-likes. Second, vibe coding and agents sit in different categories
-
What Is the AI Capability Overhang?
Ethan Mollick calls the overhang a huge gap between current AI capabilities and future capabilities. Debate still centers on what future models will do, while the capabilities of existing models are barely used and often not well understood. Better models are not required for a complete transformation of a lot of how people work in the next 10 years.
Getting up the adoption curve is the missing work. Imagination for the transformation is lacking, which is why there is no playbook yet. The change is not an automated process. The details still have to be figured out, but the capability is already in place. Transformation will happen, and the task is guiding it in a good direction.
-
How Did a $200 Million Data Migration Drop to $20 Million?
Ethan Mollick says no single move unlocks the full potential of AI inside a company. The work is structural change, and for many organizations it takes years.
A common pattern is pulling existing teams apart. A company takes its best senior engineer, puts that engineer with a subject matter expert and a marketing or strategy person, and gives the group impossible projects. One company he spoke with had a $200 million plan to migrate old mainframe data. A few months ago the company renegotiated the plan to $20 million. The CTO is thinking about doing the migration for $2 million in a couple of months. The shift means pulling teams apart, deciding what the company wants to do, and starting pr