Voice Tables: a voice-native alternative to Excel and Google Sheets

#ai #webdev #productivity

Voice Tables is an agentic AI workspace you control with your voice. You describe what you need, a CRM, a project tracker, an inventory sheet, and it builds the tables, documents, and data structure for you. We built it at Inithouse as a 3-in-1 workspace combining tables, docs, and an AI chat, with the goal of getting from a spoken idea to a working workspace in about 60 seconds.

This post looks at where Voice Tables sits relative to Excel and Google Sheets, what it does differently, and who it works best for.

What Excel and Google Sheets expect from you

Spreadsheets are flexible. That flexibility comes with a cost: you set up the structure yourself. A typical workflow looks like this:

Open a blank sheet
Name the columns, guess at data types
Write formulas for anything computed
Build a second sheet for a different view of the same data
Maintain it all manually as requirements shift

For developers, this is fine. For someone tracking jobs between client sites, invoicing while driving, or logging inventory in a warehouse, the manual setup is the bottleneck. Their hands are busy. Their laptop is in the truck.

What Voice Tables does instead

Voice Tables takes a different entry point. You open the app, tap the microphone, and say what you need: "Create a table for tracking plumbing jobs with columns for client name, address, date, status, and price."

The system uses Whisper for transcription and an LLM pipeline to parse intent and generate structured data. What comes back is a ready table with typed columns, not a blank grid.

The same voice input works for documents and for querying existing data. "Show me all jobs marked overdue" or "Write a summary of this week's completed work" run through the same pipeline. Over 30 voice commands cover creation, editing, filtering, and export.

Here is what the stack looks like in practice:

Feature	Excel / Google Sheets	Voice Tables
Input method	Keyboard, mouse	Voice, keyboard, or both
Setup time	Manual (you build the structure)	Seconds (describe it, it builds)
Data types	Manual per cell/column	Auto-inferred from description
Documents	Separate app (Word/Docs)	Built in, same workspace
AI queries	Limited (Copilot/Gemini add-ons)	Native chat, same data context
Collaboration	Real-time	Real-time
Offline	Desktop only (Excel)	Supported

Where this makes sense

The people who get the most out of Voice Tables tend to be in hands-busy roles. Craftsmen tracking job sites. Sales reps logging calls between meetings. Fitness coaches recording client progress during sessions. Event planners updating vendor lists on the go.

These are people who need structured data but rarely sit at a desk to enter it. The voice-first approach removes the keyboard dependency; the AI structuring removes the spreadsheet setup time.

For developers, Voice Tables is interesting as architecture. The Whisper-to-LLM-function-calling pipeline is a clean example of voice-to-structured-data conversion. If you are building anything with voice input that needs to produce typed, relational data (not just transcription), the pattern is worth looking at.

What this is not

Voice Tables is not a spreadsheet replacement for power users running complex financial models or pivot tables across 100,000 rows. Excel's calculation engine and Google Sheets' ecosystem of add-ons serve those cases well.

The AI layer is scoped to workspace operations: creating tables, populating them, querying data, and generating documents from that data. It stays focused on the build-and-organize loop rather than trying to be a general-purpose assistant.

The Inithouse context

Voice Tables is one product in the Inithouse portfolio. We build and test multiple MVPs in parallel, each searching for product-market fit. Other products in the studio include Ziva Fotka (an AI photo animator that has processed over 10,000 photos) and Be Recommended (AI visibility monitoring for brands). Each product ships fast, measures traction, and iterates based on real usage data.

Voice Tables is in beta. We are actively measuring voice command completion rates, workspace creation patterns, and which user segments return after the first session.