Start with the problem, not the folders
Shared drives often become difficult to navigate because different people use different naming habits. One person saves a project brief under a client name; another puts it in a general planning folder. An AI tool can help suggest consistent labels, but it cannot know your team’s intended filing rules unless you define them. Treat the project as a way to make files easier to find, not as an excuse to reorganize everything at once.
Choose a small, representative area for a pilot, such as one team’s active project folder. Ask the people who use it what they need to find and which files should stay where they are. Before making changes, note existing access permissions, shared links and folder ownership. Moving a file may change who can find or access it, depending on how the drive is configured, so keep moves manual and check their effects.
Create labels people can apply consistently
Build a short label list based on useful distinctions. A small team might label files by document type, such as Brief, Contract, Meeting Notes or Deliverable, and by status, such as Draft, For Review or Final. Keep labels separate from folders: a label describes a file, while its location determines where people encounter it. Avoid labels that depend on guessing sensitive details or a document’s legal significance.
Write a one-sentence rule for each label and include an “Unclear” option. For example, “Final” might mean the version the team has agreed to use, rather than the newest file or one with “final” in its name. Decide whether a file can have more than one label and who resolves disagreements. These rules give both the AI system and reviewers a consistent basis for decisions.
Ask AI for suggestions, not decisions
Start with the least information the tool needs. File names, file types and existing folder paths may be enough to suggest some labels. If the results are too ambiguous, consider short excerpts only for files your team is permitted to process with that tool. Check its data-handling settings and your organization’s rules before supplying document contents. Do not paste confidential files into an unapproved service just to speed up sorting.
A useful prompt defines the allowed labels, explains their rules and asks for an output that is easy to review. For example: “For each file in this list, suggest one document-type label from Brief, Contract, Meeting Notes, Deliverable or Unclear. Use only the information provided. Give a short reason and mark ambiguous cases Unclear. Do not suggest moving files.” Requesting a reason helps reviewers spot weak guesses, but it does not make a suggestion correct.
Record suggestions in a review sheet with columns for current path, proposed label, reason, reviewer decision and any proposed destination. Keep the original file path in the sheet so a reviewer can find the file and verify what it contains. If the AI tool invents details or infers a status from a suggestive filename, reject that suggestion rather than trying to make the file fit it.
Review labels before planning manual moves
Have someone familiar with the work inspect suggested labels against the actual files. Prioritize unclear cases, similarly named versions and documents whose permissions matter. Reviewers can accept a label, replace it or leave the file unlabelled. A brief note on recurring mistakes can reveal that a label rule needs improvement; revising the rule is often more useful than repeatedly correcting the same error.
Only after labels are checked should the team consider location changes. A label does not necessarily imply a move: a document can be correctly labelled and still need to remain in its current folder for an established workflow. For each proposed move, confirm the destination, check for duplicates or links that people rely on, and make the change manually. For a large cleanup, work in small batches and keep a simple move log with the old and new paths so mistakes are easier to trace.
Make the system easy to maintain
Test the result with ordinary tasks: can a teammate find the latest approved deliverable, distinguish it from a draft and understand where new files belong? If not, adjust the labels or folder guidance before expanding the project. A successful pilot should leave behind a short label guide and a clear route for reporting misplaced or ambiguously named files.
AI suggestions are most useful as a first pass over repetitive decisions. People still need to interpret context, protect access and decide whether a move is worth the disruption. By separating suggested labels from approved labels—and approved labels from manual moves—a small team can improve a shared drive without handing control of its structure to an automated guess.
