SIA "Chainweb Group"
Ⓒ 2026 All rights reserved
Manage cookies
Cookie Settings
Cookies necessary for the correct operation of the site are always enabled.
Other cookies are configurable.
Other cookies are configurable.
Context and Goal
The client is a condominium management company handling daily operational communication with residents, suppliers, and property stakeholders. The team manages a high volume of emails and relies on Google Drive for storing invoices, meeting minutes, building regulations, technical documents, and historical communications.
The goal was to reduce time spent searching for the right information and improve response quality, while keeping full human control over sensitive replies.
The goal was to reduce time spent searching for the right information and improve response quality, while keeping full human control over sensitive replies.
The Challenge
The main issue was not writing emails. The real bottleneck was retrieving accurate context quickly.
Most incoming requests required:
Most incoming requests required:
- Finding specific documents in Drive (invoices, minutes, regulations, notices)
- Checking previous email threads with the same sender
- Identifying the correct condominium based on partial references (address, unit number, internal code)
- Preparing a safe, professional response without missing details or making assumptions
The Solution
A human-in-the-loop assistant was implemented using Google Workspace integrations and an LLM (Gemini / Vertex AI). The system supports the office team by analyzing incoming emails, searching Gmail and Drive for relevant context, and preparing response drafts.
Emails are never sent automatically. Drafts are created for review and approval by the team.
Key components:
Gmail integration for reading incoming messages, classification, and draft generation
Drive integration for searching and reading relevant files, even across non-linear folder structures
Entity extraction to identify addresses, invoice numbers, building names, and key references
Safety rules to exclude newsletters, promotions, and no-reply sources
How Search Works
Instead of relying on rigid spreadsheets or manual indexing, the assistant performs intelligent search directly across:
- Drive folders and documents
- Historical email threads
- Attachments and recurring document patterns
Typical scenarios:
- A supplier requests payment details
> The assistant retrieves invoices, past conversations, and related files.01 - A resident requests a specific document
> The assistant locates it in Drive and proposes the correct link or attachment.02 - A request is unclear
> The assistant prepares a draft reply with 1 to 2 targeted clarification questions.03
Results
The assistant became a practical operational tool, acting like a digital colleague that retrieves context and prepares ready-to-review replies.
Key benefits:
- Significant reduction in manual search time
- More consistent and professional communication
- Fewer errors, with confirmation prompts when information is incomplete
- Scalable workflow applicable across multiple condominiums and folder structures
Technologies Used
Google Workspace (Gmail, Drive)
Google Apps Script automation
Gemini API / Vertex AI
Prompt engineering and security rules
Human-in-the-loop workflow design
FAQ – AI Development
No. It prepares draft replies in Gmail for review. Messages are only sent after human approval.
The assistant extracts key entities from the request (address, internal codes, supplier name, invoice number) and matches them against Drive files and email history.
The setup supports OpenAI, Google Gemini, Azure OpenAI, Anthropic Claude, and self-hosted models when required. The choice depends on data sensitivity, use case, and cost structure.
Yes. It can search across Drive folders and retrieve relevant files, including invoices, meeting minutes, regulations, and historical documents.
Yes. The workflow includes security filters, access control, and rule-based exclusions (newsletters, no-reply sources, promotions). Draft-only mode also reduces risk.
Yes. The same approach works for operations-heavy businesses such as real estate, facility management, legal offices, accounting firms, and customer support teams.
Ready to apply this to your workflow?
We work with leading AI providers and models, selecting the right technology based on the use case, data sensitivity, and cost structure:
Let’s design your Google Workspace assistant
Share a few details and get a clear implementation approach, timeline, and model options.
By clicking the button you agree to our Privacy Policy