· Mar 31, 2026
Ethan Mollick argues in One Useful Thing that AI interfaces shape access
Markdown · JSON · Original source
’s piece was published on March 31, 2026. He argues that many people cannot make full use of AI because chatbots are a poor interface for complex work, not simply because the models lack capability. He describes research in which financial professionals gained productivity on a complex valuation task, while chatbot responses also imposed cognitive load; less experienced workers were hurt most. Mollick contrasts that experience with tools designed for specific tasks and agents that work with people’s files and applications through familiar messaging interfaces. His examples include Cowork with Dispatch, which he says can handle complex work but remains error-prone, limited, and subject to security risks. He argues that AI-generated interfaces could adapt to the task rather than requiring people to adapt to one fixed interface. His conclusion is that better interfaces could let more people use existing AI capabilities, making the technology feel more capable even without a model change. Mollick predicts that this will change soon, and suspects that some “AI disappointment” stems from interfaces being wrong. The evidence he offers includes a small-group study and his own examples; it does not establish how broadly these experiences generalize.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Mar 31, 2026 (2026-03-31T22:34:37.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-07T17:03:23.637Z. First observed: 2026-09-23T05:37:00.454Z. Retrieved: 2026-10-24T06:47:35.193Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 24, 2026; originally published March 31, 2026. The described valuation research involved a small group of financial professionals; the text does not establish how broadly its findings apply. Mollick’s examples of Cowork and Dispatch are personal demonstrations, not evidence of performance across users or tasks. The article’s predictions about future interfaces and broader AI access are forecasts, not reported outcomes.
Open question
How much can better interfaces improve people’s ability to use AI for complex work, especially for less experienced workers?
Source-stated
Ethan Mollick argues that chatbot interfaces can obstruct complex work by imposing cognitive costs, even when AI provides productivity gains; he says less experienced workers were hurt most in the study he describes.
People did see a productivity gain from using AI, but some of that seemed to be offset by the fact that the AI presented information in a way that completely overwhelmed people
Source-stated
Mollick argues that familiar messaging interfaces let people direct an AI agent to work with their files and tools in a more natural way.
People don’t want a chatbot. They want an agent that works on their actual files, with their actual tools, accessible the way they talk to people.
Source-stated
Mollick says Claude Cowork with Dispatch can perform complex work, but is limited and error-prone, and still carries security risks.
Cowork is sandboxed, safer but more limited (but that doesn’t mean there aren’t security risks). The connector ecosystem is growing but incomplete.
Source-stated
Mollick predicts that AI will generate interfaces suited to the task and argues that better interfaces could let more people use AI capabilities.
I suspect the future isn’t one interface to rule them all. It’s AI that generates the right interface for the moment
Source-stated
Mollick suspects some AI disappointment comes from unsuitable interfaces rather than poor AI, and predicts that this will change soon.
My guess is that a lot of the “AI disappointment” people sometimes express comes not from the AI being bad, but from the interfaces being wrong.
Source and contributor context
One Useful Thing · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Ethan Mollick
Ethan Mollick is a Wharton researcher and author exploring how AI changes work, learning and decision-making.
One Useful Thing
One Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Editorial provenance
Version 1. Publication ID: 899436a6-8062-edf1-18c0-ee9405a8dd9a.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-07T16:58:47.848Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: a4dcc994bd51c962c1e510a3c879174cbdec0f47f5926f2e7f4058705b2ff42b. Source text hash: 0172675986eb819d5a698eba3ce9b401f42dddbd1f252ff0c0c217457f2c94d7.
Revision and source-check history
Version 1 · 2026-11-07T17:03:23.637Z
Ethan Mollick argues in One Useful Thing that AI interfaces shape access
’s piece was published on March 31, 2026. He argues that many people cannot make full use of AI because chatbots are a poor interface for complex work, not simply because the models lack capability. He describes research in which financial professionals gained productivity on a complex valuation task, while chatbot responses also imposed cognitive load; less experienced workers were hurt most. Mollick contrasts that experience with tools designed for specific tasks and agents that work with people’s files and applications through familiar messaging interfaces. His examples include Cowork with Dispatch, which he says can handle complex work but remains error-prone, limited, and subject to security risks. He argues that AI-generated interfaces could adapt to the task rather than requiring people to adapt to one fixed interface. His conclusion is that better interfaces could let more people use existing AI capabilities, making the technology feel more capable even without a model change. Mollick predicts that this will change soon, and suspects that some “AI disappointment” stems from interfaces being wrong. The evidence he offers includes a small-group study and his own examples; it does not establish how broadly these experiences generalize.
Original source
· Article
Source details
Original source date: Mar 31, 2026 (2026-03-31T22:34:37.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-07T17:03:23.637Z. First observed: 2026-09-23T05:37:00.454Z. Retrieved: 2026-10-24T06:47:35.193Z.
Limits
Text as retrieved on October 24, 2026; originally published March 31, 2026. The described valuation research involved a small group of financial professionals; the text does not establish how broadly its findings apply. Mollick’s examples of Cowork and Dispatch are personal demonstrations, not evidence of performance across users or tasks. The article’s predictions about future interfaces and broader AI access are forecasts, not reported outcomes.
Open question
How much can better interfaces improve people’s ability to use AI for complex work, especially for less experienced workers?
Source-stated
Ethan Mollick argues that chatbot interfaces can obstruct complex work by imposing cognitive costs, even when AI provides productivity gains; he says less experienced workers were hurt most in the study he describes.
People did see a productivity gain from using AI, but some of that seemed to be offset by the fact that the AI presented information in a way that completely overwhelmed people
Source-stated
Mollick argues that familiar messaging interfaces let people direct an AI agent to work with their files and tools in a more natural way.
People don’t want a chatbot. They want an agent that works on their actual files, with their actual tools, accessible the way they talk to people.
Source-stated
Mollick says Claude Cowork with Dispatch can perform complex work, but is limited and error-prone, and still carries security risks.
Cowork is sandboxed, safer but more limited (but that doesn’t mean there aren’t security risks). The connector ecosystem is growing but incomplete.
Source-stated
Mollick predicts that AI will generate interfaces suited to the task and argues that better interfaces could let more people use AI capabilities.
I suspect the future isn’t one interface to rule them all. It’s AI that generates the right interface for the moment
Source-stated
Mollick suspects some AI disappointment comes from unsuitable interfaces rather than poor AI, and predicts that this will change soon.
My guess is that a lot of the “AI disappointment” people sometimes express comes not from the AI being bad, but from the interfaces being wrong.
Source and contributor context
One Useful Thing · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Ethan Mollick
Ethan Mollick is a Wharton researcher and author exploring how AI changes work, learning and decision-making.
One Useful Thing
One Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Editorial provenance
Version 1. Publication ID: 899436a6-8062-edf1-18c0-ee9405a8dd9a.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-07T16:58:47.848Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: a4dcc994bd51c962c1e510a3c879174cbdec0f47f5926f2e7f4058705b2ff42b. Source text hash: 0172675986eb819d5a698eba3ce9b401f42dddbd1f252ff0c0c217457f2c94d7.
uncertain · 2026-11-07T16:58:47.848Z · anthropic/claude-opus-5.5
Publication: 899436a6-8062-edf1-18c0-ee9405a8dd9a
attribution — pass: The reading attributes the essay to Ethan Mollick in One Useful Thing, matching the byline and publisher. It presents the arguments, including the claims about research findings, as his assertions.
Claude Dispatch and the Power of Interfaces We often lack the tools for the job, even if the AI is capable enough Ethan Mollick
summary — uncertain: The summary is mostly faithful and readable. One passage is ambiguous: 'Mollick predicts that this will change soon' follows the sentence about AI feeling more capable 'even without a model change,' so a reader could think the model is what will change. Mollick means that making people access AI by typing into a chat window will change soon. Smallest repair: 'predicts that reliance on typing into chat windows will change soon.' Optional addition: noting that Cowork and Dispatch are Anthropic tools, and that Dispatch lets a user message the desktop agent from a phone.
We built one of the most powerful technologies in recent history and then made people access it by typing into a chat window. That will change soon.
date — pass: originalPublished is 2026-03-31T22:34:37Z, and its UTC calendar day matches the stated March 31, 2026 and the Mar 31, 2026 dateline. Predictions are framed as Mollick's forecasts. The retrieval date is disclosed. The evidence kind is a publisher page, not an archive copy, and there is no hindsight framing.
Ethan Mollick Mar 31, 2026
support — pass: Each claim is supported by the quoted text: the productivity gain offset by cognitive load, the harm to less experienced workers, agents working on actual files and tools, Cowork's limits and security risks, AI-generated interfaces, and interface-driven disappointment. The limitations correctly note the small study group and Mollick's personal demonstrations. The implied benefit to less experienced workers stays conditional, as it is in the source.
The people hurt most were less experienced workers, exactly the people who could benefit the most from AI… if they could keep track of what they were doing with it
media — pass: No media was selected, and no verified media candidates were supplied, so there is no media identity or relevance issue.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
One Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice.
Explore One Useful ThingEthan Mollick is a Wharton researcher and author exploring how AI changes work, learning and decision-making.
Explore Ethan MollickClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeOne Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice.
Explore One Useful ThingClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeEthan Mollick is a Wharton researcher and author exploring how AI changes work, learning and decision-making.
Explore Ethan MollickClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeOne Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice.
Explore One Useful ThingClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore Claude