Brief IA
›

Dropbox: 1,200 Images Organized in 10 Minutes

Business & Startups·Tom Levy·

Dropbox: 1,200 Images Organized in 10 Minutes

Dropbox: 1,200 Images Organized in 10 Minutes
Key Takeaways
1According to Dropbox, Randall Barney sorted 1,200 images in about 10 minutes with his new experience.
2Dropbox files can be used in ChatGPT, Claude, Copilot, and Muse, with the results then saved in Dropbox.
3Mike Torres recommends setting access and modification rights for agents.
Why it matters : Mike Torres advocates for evaluating AI based on the final outcome, its quality, and the feedback from the people doing the work.

According to Dropbox, Randall Barney, CEO of Intelligent Communities Group, classified 1,200 images in about 10 minutes using the new service experience. Mike Torres, Dropbox's product director, advocates for an AI evaluation based on completed work rather than just the speed of generation.

Two Clients Report Dropbox Use Cases

Randall Barney then added the resolution of each image to its filename. Dropbox also cites the creative agency YYES: reorganizing a project folder reportedly took two or three minutes with this experience, compared to a usual estimated time of 15 to 20 minutes.

However, these durations are only part of the evaluation recommended by Mike Torres. He suggests first measuring the time spent on the existing process, its manual operations, delays, and the steps that need to be repeated. The quality of the outcome and the smooth transition to the next steps should also be considered. Producing a draft in a few seconds is of little value if its review or preparation for the recipient then takes hours.

The criterion he proposes focuses on the completion of collective work: reducing difficulties without compromising quality requirements. He gives equal weight to feedback from those doing the work as to quantitative measures, as their experience can indicate the improvements achieved and the adjustments still needed. Success, in his view, is also measured by the ability to complete a useful task while simplifying the work of others.

Usable Dropbox Files Across Multiple AI Tools

According to the company, Dropbox's new experience relies on the context contained in the files that users already own to help them advance their projects. Mike Torres advocates for the freedom to choose one's AI tool without relinquishing control over content.

Users can work on their Dropbox files in compatible tools, including ChatGPT, Claude, Copilot, and Muse, and then save the results back to Dropbox. The company presents this shared storage as a means to submit content for review, rework it, and share it within teams, even when needs or applications change.

La lettre du soirGratuit · 19 h

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Information Dispersed Across Multiple Tools

AI is already present in corporate systems, while many companies continue their experiments. Mike Torres identifies the daily obstacle as the difficulty in leveraging information that is available: accumulated content is scattered across folders, tools, and systems. Finding the relevant document, understanding its context, verifying its version, and then reusing it can take more time than expected. When repeated within a team, these interruptions add to the workload.

Research, file transfers, version control, and administrative tasks then occupy a portion of the time that could be spent using this information. Collaboration also suffers when colleagues do not always have the same context or necessary access. Specialized tools have facilitated storage, exchanges, project management, or sharing, but a single project can still distribute documents, conversations, comments, and deliverables across multiple applications. Content remains available without necessarily following its context.

For Mike Torres, simply adding an application does not necessarily solve the problem. AI can help leverage various formats without requiring perfect organization or a single location. The goal he advocates is to make existing resources easier to use, allowing more room for judgment and expertise from individuals.

A summary or draft, even if useful, often covers only a fraction of the task. To continue the work, it is necessary to connect AI to the relevant elements and make its output usable in the subsequent process. This requires considering their location, permissions, reviewing outputs, and existing practices. Mike Torres thus recommends starting from the concrete problem to be solved and the desired outcome.

Defining Authorized Actions for Agents

Mike Torres distinguishes between consulting an agent and granting it permission to intervene. To begin, he advocates for precisely defined repetitive tasks, with an explicit expected outcome and verifiable execution. Classifying information, moving from one file to another, or chaining routine operations are among the uses he deems suitable for automation.

According to him, companies must specify the resources accessible to the agent, the modifications it can make, and the situations in which a person must intervene.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.