Business
Co-Intelligence
Also known as: human-AI collaboration
The idea of working *with* AI as a collaborator and thinking partner, rather than treating it as either a magic oracle or a mere tool.
Co-intelligence is the idea that the best way to use AI is with it, as a partner you work alongside, not as a genie that does everything for you or a gadget you barely touch. You bring judgement, goals and checking; the AI brings speed, breadth and tireless drafting. The magic is in the collaboration: you get more done and often think better, as long as you stay involved and verify.
Co-intelligence describes a working relationship in which humans and AI complement each other. The framing pushes back on two extremes: treating AI as an infallible oracle (dangerous, because it hallucinates) or dismissing it as useless. Reality sits in between, a capable but flawed collaborator.
Practical principles (in this spirit)
- Always invite AI to the table: try it on your real tasks to learn where it helps.
- Stay the human in the loop: you own judgement, ethics and the final call.
- Treat it like a person, but remember it is not one: interact naturally, yet verify.
- Assume this is the worst AI you will ever use: it keeps improving, so build the habit now.
This is closely tied to augmentation: using AI to amplify people rather than simply replace them.
Co-intelligence (Mollick 2024) is a framing for human-AI collaboration that rejects both over-trust and dismissal, positioning current models as broadly capable but unreliable partners whose value is realised through active human direction and verification. It reframes the unit of performance from "AI alone" or "human alone" to the human-AI system, aligning with evidence that well-designed collaboration can raise output and quality while poorly-designed reliance degrades both (automation complacency, deskilling).
Operationally it emphasises human-in-the-loop oversight, task-level judgement about where models help ("jagged frontier": strong on some tasks, surprisingly weak on adjacent ones), and AI literacy built through hands-on use. It connects to the augmentation-versus-automation debate and to workflow design questions, when to let AI draft, critique, or act, and where to insert verification. As a practitioner concept rather than a technical mechanism, its rigour comes from empirical studies of human-AI teaming and from disciplined verification practices that counter hallucination and bias.