Claim
AI labs have crossed the threshold they always stated as their goal: building autonomous systems that outperform humans at most economically valuable work. This renders the co-intelligence frame obsolete and requires a new frame, co-existence, for working alongside systems that are "sometimes, but not always, better than you."
Mechanism
The co-intelligence frame, which Mollick described in his 2024 book, was suited to a world of chatbots and earlier models where working with AI was a cooperative exercise: the human directed, the AI assisted, and the human remained at the center. The new frame, co-existence, is needed because the capability relationship is now asymmetric and variable. AI systems can outperform humans at many economically valuable tasks, but not all. The difficulty shifts from "how do I collaborate with AI?" to "how do I identify and act on the situations where I am better, and where AI is better?" The frame also changes the stakes: Mollick notes that AI readers will increasingly function as "the gatekeeper standing between your work and its audience," which means the human relationship with AI is no longer primarily collaborative but also competitive and evaluative in both directions.
Conditions
Holds when: the AI system in question has demonstrated the ability to outperform humans on a meaningful share of the task class in question.
Fails when: AI capability in the relevant domain has not yet crossed a meaningful human-performance threshold, making the co-intelligence cooperative frame still appropriate.
Evidence
Mollick wrote this two years after publishing Co-Intelligence, noting the original book was "written about a world of chatbots and earlier AI models where working with AI was a cooperative exercise of prompting back-and-forth." Recent AI coding benchmarks show models writing 80% of code at some organizations, with teams shipping 8x more than before. He notes the threshold that labs always stated as their explicit target has now been reached:
"Their goal has always been, for better or worse, to build what the OpenAI charter calls 'highly autonomous systems that outperform humans at most economically valuable work.'"
Signals
- You can no longer assume that your first draft, analysis, or proposal is better than what an AI would produce unaided for the same task.
- The relevant question in your workflow shifts from "how do I use AI to improve my work?" to "in which parts of this task am I reliably better, and which parts should I hand off?"
Counter-evidence
Mollick does not claim AI is uniformly better than humans. The co-existence frame depends on the variability: AI is "sometimes, but not always, better." Teams in domains where AI capability is still narrow or unreliable may find the co-intelligence frame more useful than the co-existence frame for most of their daily work. The threshold is also not a sharp line: capability varies significantly by task type, domain, and the specific model in use.
Cross-references
- Frontier AI shifts the human role from wizard who steers through prompting to patron who commissions outcomes and judges the result: The practical, task-level expression of this framework shift. If co-existence is the macro frame, the patron dynamic is what it looks like in a single workflow.