Source: Fügener, Walzner & Gupta (2026) — “Roles of Artificial Intelligence in Collaboration with Humans: Automation, Augmentation, and the Future of Work,” Management Science
The AI-at-work debate often gets trapped in a very boring argument.
Will AI replace people?
Will AI assist people?
Will AI create a magical workplace where everyone is productive and nobody has meetings that could have been emails?
This paper is useful because it refuses the simple answer. It says the real question is task allocation.
Some tasks should be automated. Some should be augmented. Some should stay with humans.
That sounds obvious, but organisations are very talented at making obvious things complicated.
The authors distinguish between two kinds of complementarity. Between-task complementarity means AI is better at some tasks and humans are better at others. In that case, automation makes sense because AI can take the easier or more AI-friendly work, freeing humans for harder tasks. Within-task complementarity means humans and AI perform better together on the same task. In that case, augmentation makes sense.
The paper validates this with image-classification data. The pattern is elegant: AI automates relatively easy tasks, AI assists humans on tasks where their performance is similar, and humans work alone on the hardest tasks.
I like this because it sounds much more realistic than “AI everywhere.”
Not every task needs a human-AI collaboration layer. Sometimes AI should just do the thing. Sometimes the human needs advice. Sometimes the human needs quiet, context, and freedom from another dashboard pretending to help.
The hidden lesson is resource reallocation. Automation only creates value if the humans freed from routine work are moved to work where they actually add value.
Otherwise, AI just creates spare capacity that disappears into more meetings. Tragic but believable.
My takeaway: the future of work is not about choosing automation or augmentation as a slogan. It is about knowing where each one belongs. The hard part is not the model. It is the job design.