The AI Productivity Paradox: Why Aren’t More Workers Using ChatGPT?

The article argues that ChatGPT and other LLMs can do more than summarization, but most knowledge workers still barely use them. The main barrier is not the tools themselves; it is the way organizations prioritize urgent delivery over time for experimentation, workflow review, and deeper rethinking.

It also says AI gains often come from focused, hands-on use, not one-click automation. Short, domain-specific training, internal AI enthusiasts, and scheduled time for exploration can help teams find higher-value uses and identify where AI can improve work.

Reading notes

#

• Most knowledge workers the author speaks with do not use ChatGPT at all, and those who do mostly use it for summarization.
• A small share of users pay for ChatGPT Plus, which the author treats as a sign that few people are using it for complex professional work.
• The author says the real obstacle is how organizations approach work, not the potential of AI.
• The right executive question is how AI can create more value, not only how it can make tasks faster.
• In one day, the author used Google AI Studio to turn unstructured data into a structured dataset, identify user groups, and build a new taxonomy.
• That result required detailed prompts, feedback, and several iterations, not simple automation.
• The author says the process compressed about a month of work into a day, but it was mentally exhausting.
• The benefit was not only speed; LLMs exposed patterns and edge cases that structured analysis would have missed.
• The key condition for that success was having leadership support to spend a full day rethinking data processes with AI.
• Most PMs and other workers lack time for exploratory work because they are under pressure from deadlines, customer requests, and layoffs.
• This pressure also makes AI adoption for better execution harder, because testing and issue-finding are treated as luxuries.
• The author argues that many people need some training, but not the heavy or highly technical kind often sold in the market.
• Short, tailored workflow audits of 10 to 15 minutes can be more effective than generic AI classes.
• The author rejects the idea that AI adoption is limited to technical workers under forty.
• Attention to detail and care for good work matter more than technical background.
• The author gives the example of his father, a lawyer in his sixties, who understood LLMs quickly once the examples were tailored to his domain.
• The company may already have internal AI enthusiasts who can help others learn through workflow audits and starter prompts.
• Employees also need time to explore and experiment in their own domain after they understand the tools.
• The conclusion is that leadership must create space for open-ended, goal-driven work if it wants AI adoption to change how people work.