When Avatar became the highest-grossing film of all time, critics had a consistent complaint: the story was deeply familiar. Reviewers joked that if you’d seen Dances with Wolves or Pocahontas, you’d already seen Avatar. James Cameron borrowed his plot freely. The critics saw laziness, but it was a strategy. So much about the film was radically new that Cameron made whatever he could feel familiar so audiences would not be overwhelmed.
In my research on the creative process, published in the Academy of Management Review, my coauthor and I called this novelty reduction: the deliberate practice of dialing down newness in some parts of a creation so the genuinely new parts can succeed. It runs against the instinct that more creative is always better. Often the most successful innovations are the ones that hide how new they really are.
I recently argued that in the world of AI, the returns go to whoever finds a great director, not whoever buys a better camera. Novelty reduction is one of the things great directors do, and it is a critical lesson that enterprise AI has not yet learned.
One reason ambitious AI projects stall is that too much novelty arrives at once: a new interface, a new workflow, a new way of making decisions. All these changes land on people who feel like they were doing fine yesterday. Novelty is not free. It is a cost the user pays in attention, effort, and risk. Pile on enough of it and even a brilliant system gets quietly ignored. The gatekeepers resist novelty for the same reason studios resist strange scripts: unfamiliar things are hard to evaluate, so they get rejected. And gatekeepers are exactly the managers, buyers, and veterans whose buy-in you need.
The fix is to spend your novelty budget where it counts and reduce it everywhere else. There are two ways to do that.
Permanent novelty reduction
Permanent novelty reduction means stripping newness, for good, from the parts that don’t need it. The intelligence should be novel; the experience around it shouldn’t be. A salesperson about to walk into a meeting doesn’t want a new platform to master. They want a short brief that reads like the notes a sharp colleague would have handed them. The AI underneath can be sophisticated. What the user touches should feel like something they’ve done a hundred times. Making novelty invisible is what lets people adapt.
Temporary novelty reduction
Temporary novelty reduction means starting more familiar than you eventually intend to be. Resistance to a new tool is highest before anyone trusts it. So you open with a modest, recognizable on-ramp: a capability people immediately understand. You earn their confidence, then expand toward the more ambitious version once the relationship can bear it. The full vision still arrives, just not on day one. Cameron earned the right to show you a floating mountain by first telling you a story you’d already heard.
Put the newness where the value is
Novelty reduction reframes what good AI adoption looks like. The goal is not to impress people with how futuristic the system is. It is to concentrate the newness in the value: the quality of the insight, the time saved, the decision improved. The path to that value should feel almost ordinary. The best implementations don’t announce a revolution. They feel like a slightly better version of Tuesday.
Avatar was a technological leap that audiences experienced as a familiar story told beautifully. The enterprises that win with AI will do the same: radical capability, delivered in a form that barely asks the organization to change at all. The novelty that matters most is the kind your users never have to think about.
Reference: Katz, J. H., & Ellis, L. M. (2025). Dances with Avatar: How creators can reduce the novelty of their work to achieve more creative success. Academy of Management Review, 50(1), 148–159. https://doi.org/10.5465/amr.2022.0511