Hard-to-find files
Employees may know the answer exists but not which document, folder, or section contains it.
Back to case studies
Working prototype
Google Drive Knowledge Assistant
Company policies, SOPs, and reference files often already contain the right answers. The problem is that employees cannot find those answers quickly, so the same questions keep landing on one knowledgeable person. This prototype turns a shared Google Drive folder into a searchable internal reference employees can ask in plain language.
Use caseInternal knowledge
Document sourceGoogle Drive
StatusWorking prototype
Overview
Ask the company files before interrupting another person.
Employees ask a question through chat. The workflow searches the available company documents, retrieves the most relevant sections, and uses those sections to prepare a concise answer.
The business problem
The documents existed, but finding the right answer still took too long.
A shared folder can store years of policies and operating knowledge without making that information easy to use. Search depends on filenames and memory, while repeated questions continue to interrupt experienced employees.
Repeated questions
The same policy and process questions repeatedly land on managers or experienced team members.
Inconsistent answers
Different employees can interpret or remember the same policy differently.
Slow onboarding
New hires depend on other people to explain information that is already documented.
Changing documents
A useful assistant must recognize new and updated files instead of relying on a one-time upload.
Unsupported-answer risk
If the files do not contain an answer, the system should say so instead of inventing a company rule.
Design boundaries
Useful internal answers require clear source boundaries.
- Keep the existing folder workflowThe team should continue managing approved files in a familiar shared Drive folder.
- Recognize document changesNew and updated files need to enter the preparation pipeline automatically.
- Use matching source materialThe answer should be composed after retrieving relevant document sections.
- Fail honestlyIf the documents do not contain the answer, the assistant should return an explicit fallback.
- Respect document qualityOutdated, vague, or contradictory files will produce weaker answers and still need business ownership.
- Separate prototype from productionThe workflow demonstrates the end-to-end design without claiming production scale or measured business impact.
Solution
One workflow prepares the knowledge. Another retrieves it for employees.
The first pipeline watches a selected Google Drive folder. When a file is created or updated, the workflow downloads it, breaks the content into searchable sections, creates a searchable representation, and stores those sections in the knowledge index.
The second pipeline receives an employee question through chat. It searches the indexed company content, retrieves matching sections, and gives that context to the assistant before an answer is prepared.
The important boundary is simple: available documents are the reference. When the required information is missing, the assistant uses an explicit unavailable-answer response rather than pretending a policy exists.
System architecture
Document updates in. Relevant company answers out.
The implementation separates document preparation from question answering so each pipeline can be inspected and improved independently.
Shared Drive folder
Document preparation
Searchable knowledge
Employee question
Relevant sections
Grounded answer
Explicit fallback
Working proof
The prototype connects document updates to a usable question-and-answer flow.
- New and updated files detectedSeparate Drive events feed the same document preparation path.
- Documents prepared for searchDownloaded files are split into smaller sections before they are indexed.
- Questions retrieve matching sectionsThe chat flow uses a retrieval tool before the assistant prepares its response.
- Missing information has a fallbackThe assistant is instructed to state when the available resources do not contain an answer.
- Both pipelines are visibleThe supplied screenshots show the ingestion and question-answering architecture without exposing private data.
- Public demonstration publishedA short TikTok shows the business problem and the prototype concept in a practical format.
Technologies
The tools support document retrieval; the business outcome stays simple.
Google Driven8nGoogle GeminiPineconeDocument splittingRetrieval-based answeringChat interface
Workflow gallery
Two focused pipelines power the assistant.
These screenshots show the working n8n structure. They contain no client names, customer details, private URLs, credentials, or internal identifiers.
Document preparation pipeline
New or updated Drive files are downloaded, divided into searchable sections, represented for search, and stored in the document index.
Question-answering pipeline
An employee question reaches the assistant, which searches the company-document index and uses the matching sections as answer context.
Public demonstration
The working idea, explained in 27 seconds.
This TikTok presents the business problem without leading with technical terms: company answers already exist, but they are buried inside documents people struggle to find.
The video is included as public proof of how the prototype was demonstrated—not as a claim of measured client results.
Why this matters
Internal knowledge should not disappear into folders or depend on one person's memory.
A useful company assistant does not replace the documents or their owners. It makes approved information easier to retrieve while clearly admitting when the available files do not contain the answer.
Does your team keep asking questions that are already documented?
Book a free discovery call and we will identify whether your policies, SOPs, and reference files can become a more useful internal knowledge workflow.
Book My Free Discovery Call