Did you know that a study with consultants showed gains with AI, but not in all tasks? It follows seven hundred and fifty-eight BCG consultants. Its central point is the boundary between the capabilities of the tool and the tasks that exceed them. For some tasks tailored to its capabilities, the assistant can help produce and organize a response. Our work table shows comparative proposals, without inventing a real advice to a company. The person must examine the argument and context before choosing an answer.
The search version reports more tasks done, faster work and better quality on this perimeter. Numbers should not be mixed with a version. Above all, a good result within this border does not guarantee anything beyond.
The practical lesson is to recognize the tasks where help is useful and those where it can deceive. The illustration is not a result chart. An assured answer remains to be verified, even when the tool has succeeded in other exercises in this scientific study.
What we know
The research involves 758 consultants and a task-dependent capability boundary. Results from different versions should not be mixed.
To go to the source
References consulted on October 12, 2026.
Signaler une erreur ou apporter une précision
