Dropbox is considering memory and attention training as AI use expands. The evidence points executives toward questions about effort and retained judgment.
On 2026-10-25, Fast Company reported that Dropbox was beginning to consider training around brain health and cognitive well-being. Chief People Officer Melanie Rosenwasser described memory exercises and attention-span training as possible components of that effort.
The report also described experiments with device-free executive offsites that included breathwork. Rosenwasser said Dropbox wanted to understand how everyday AI use affects employees and consider ways to support critical thinking and human connection.
For founders and executives, the important development is the management question behind these activities. Dropbox is encouraging employees to use AI while asking what happens when more thinking is delegated to it. That puts efficiency and retained human judgment on the same agenda.
Dropbox’s cognitive-health discussion sits alongside an active push toward AI adoption. According to Fortune, CTO Ali Dasdan said fewer than 40% of the company’s engineering, product and design teams used AI tools when he joined the role. He said adoption across those teams subsequently reached full coverage.
Dasdan described training and AI-focused hackathons as parts of that expansion. Employee “superusers” and shared metrics also featured in his account. Fortune reported that Dropbox made ChatGPT Enterprise widely available to knowledge workers and approved AI features integrated into several workplace tools.
Employees had created hundreds of custom applications, according to Dasdan. He described AI efficiency as one consideration in project staffing rather than tying it directly to headcount. These details matter because the cognitive-health discussion is taking place inside a company already encouraging broad experimentation.
A secondary research summary describes a 2025 CHI paper by researchers from Microsoft Research and Carnegie Mellon University. The researchers surveyed 319 knowledge workers about 936 real-world generative-AI use examples.
The study examined workers’ reported critical-thinking effort and their confidence in AI and themselves. Higher confidence in AI was associated with less reported critical-thinking effort. Higher confidence in one’s own ability was associated with more.
For executives, this makes confidence a useful subject for review. The management question is how employees assess an answer when they already expect the system to be reliable. It is also worth asking whether employees understand the reasoning well enough to explain why the output deserves acceptance.
Together, the corporate reporting and workplace study point toward a work-design issue: where should human reasoning remain explicit as AI becomes more widely available?
The practical implication is to evaluate efficiency and retained judgment together. That is an editorial inference from the reporting and study findings, not a tested Dropbox procedure. It gives leaders a concrete way to examine AI deployment without assuming that tool use and cognitive dependence are interchangeable.
An adoption review asks whether people can access and use a system. A judgment review asks who remains responsible for understanding its output. Both questions belong in an executive discussion about how work should be divided between people and software.
A useful review starts with the thinking a task requires. Is the employee framing the problem? Are they comparing alternatives? Can they explain the assumptions behind the recommendation? These questions make retained judgment more concrete than a broad instruction to remain critical.
The aim is to identify where reasoning should be visible in the work product or decision discussion. A polished document and a defensible conclusion are different standards. Leaders can ask for the reasoning that connects evidence to a recommendation rather than treating presentation quality as the complete assessment.
This is also the management question behind our coverage of cognitive effort during AI-assisted tasks. The relevant issue is the role assigned to the person within the task, rather than the mere presence of an AI tool.
For work that also serves a learning purpose, leaders can examine which parts employees still need to practice themselves. That might mean discussing how someone would approach a problem before reviewing an AI-generated response. The purpose is to make independent understanding visible rather than presume it from the finished output.
Managers can also ask employees to explain why they accepted or rejected a suggestion. This is a proposed way to examine reasoning during ordinary work. It should be evaluated on whether it reveals understanding and improves the quality of the decision process.
For founders and investors assessing organizational capability, the central question is straightforward: does the team retain enough understanding to challenge the system when the stakes require it?
For demanding professional lives, sustained focus is a useful lens for these questions. Which tasks still require employees to hold context and weigh competing explanations? Where are they expected to notice an inconsistency? The answers can help define the human contribution that an organization wants to preserve.
Leaders should also distinguish the effort required by a task from the broader experience of mental fatigue in AI-heavy work. In management discussions, clearer labels help keep questions about attention, learning and output quality separate. Each deserves an appropriate assessment rather than a single catch-all score.
The limits fall into distinct categories. First, the Dropbox reporting describes concern and experimentation. It does not report measured cognitive decline among employees or provide outcomes showing that the proposed exercises or device-free offsites improve cognition. The activities should therefore be understood as organizational experiments.
Second, the workplace study relied on self-reported information and identified associations. It was not a direct or longitudinal measurement of cognitive decline. Less reported critical-thinking effort on an AI-assisted task does not establish lasting loss of ability. Effort and capability are different outcomes.
Third, the brain-training evidence requires a separate assessment. Fast Company noted that evidence for improvements in general intelligence is limited. It also cited an NIH-funded study in which reasoning and processing-speed training was associated with benefits in those specific abilities among older adults at a ten-year follow-up.
That finding concerns particular trained abilities in an older population. It does not establish that memory exercises or attention training will counteract AI-related changes among working-age employees. Success on a practiced exercise and stronger independent judgment at work need to be assessed separately.
The practical response is proportionate: examine how work is allocated and how understanding is demonstrated. Treat cognitive-training claims as questions requiring their own evidence. Where concerns involve an individual’s cognitive health, personal assessment belongs with qualified care rather than an employer’s AI-use scorecard.
The next useful evidence would follow people over time and measure abilities directly. Research could examine whether different forms of AI reliance affect independent problem-solving or performance on unfamiliar tasks. That would help distinguish a change in how effort is allocated from a lasting change in what someone can do.
Training studies would need to answer another question: do gains transfer to meaningful work? For an executive audience, useful outcomes would include the ability to identify a weak assumption or defend a recommendation. Better performance on the training exercise itself would answer a narrower question.
Future evaluations could also distinguish the components of a workplace initiative. Memory exercises and attention training should be assessed on their stated objectives. Device-free offsites should likewise be evaluated against clearly defined goals rather than grouped under a general promise of better cognition.
For employers, the immediate opportunity is to broaden the definition of successful AI adoption. Usage and efficiency can sit alongside questions about learning and independent judgment. This does not require resolving the entire science of AI and cognition before reviewing how a team works.
Rosenwasser told Fast Company that Dropbox was approaching the subject with “curiosity and humility.” That is an appropriate stance for an emerging question. The next phase should bring clearer definitions and better measurement of the human capabilities organizations want to preserve.
Stay connected for research and practical guidance on executive performance, energy, focus, sleep, recovery and longevity. Ideas built for people who want to stay sharp, capable and effective for the long run.
Build habits and systems that support clear thinking, steady energy and long term capacity throughout a demanding career.
explore the Blog
