Dropbox’s AI survey puts the shared folder in focus

The file still needs somewhere to go

Dropbox published its 2026 report on AI and work content on 14 October, drawing on a survey of 219 professionals at The AI Conference in San Francisco. Its figures describe that group’s reported habits. One practical question stands out for us: once an AI-assisted draft is useful, can the next person find it? [1]

Respondents estimated that AI created or substantially changed an average of 46.4% of the new work files they created or saved. That is an average estimated share of files, not a finding that 46.4% of all workers use AI. It is also self-reported survey data, rather than a count taken from file systems. [1]

A destination is part of the task

In the report’s list of destinations for useful AI output, 45.2% of respondents named a shared team workspace and 44.3% named a new personal or local file. The report describes multiple destinations, so those categories should not be read as mutually exclusive shares of all output. [1]

Our interest is the handover between creating something and making it available to someone else. A clear explanation saved in a personal folder may still need a name, a home and a little context before a colleague can use it. Producing another draft does not settle those choices.

Read the sample alongside the headline

The American Association for Public Opinion Research’s best-practice guide explains why sample design and transparent reporting matter. It recommends describing who was studied, how participants were recruited, the questions and how results were analysed. The linked guide was updated in March 2022; it is methodological context, not a review of Dropbox’s findings. [2]

For that reason, we would keep this report’s conference setting attached to its numbers. A group encountered at an AI event is not established here as a representative sample of South African businesses. We have not independently reproduced the survey or measured the workflows it describes. The report can prompt useful questions without supplying a national benchmark. [1] [2]

Make the handover visible

For a small South African team, our suggested exercise would be to follow one ordinary, non-sensitive project brief from draft to handover. Agree where the reviewed version belongs, give it a recognisable name, and leave a short note explaining the sources and decisions behind it. Then ask a colleague to find the right version without help from its author. This is an exercise we propose, not one we have performed.

That modest check keeps the discussion close to the work. Could someone locate the document, understand its status and carry on from it? Our view is that a useful AI workflow needs an answer to those questions as much as it needs a good first response. The shared folder deserves some attention of its own.

AI-assisted research and writing, with a separate source-checking pass by the same Codex assistant. Dropbox supplies self-reported conference survey findings; AAPOR supplies historical methodology guidance, not corroboration of those findings. We have not replicated the survey or performed the proposed handover exercise. No local adoption, financial savings, causal productivity or employment claim is made.