Academy/Integrations/Connect a Google Sheet
GuideIntegration

Connect a Google Sheet

Give your agent a living lookup table. Keep your prices, listings, or stock in a Google Sheet, and the agent searches it in real time to answer — you update the sheet, the answers update with it.

AAssistantLabs Team
Academy
6 min read Updated Jul 2026

When to use a sheet

A Google Sheet is the right home for data that’s big or changes often — a price list, a stock table, apartment listings, store locations, a menu. Connect it once, and your agent searches it live every time a customer asks. You keep working in the spreadsheet you already have; the agent reads from it.

Three places knowledge can live — pick by the shape of the data:

  • A Google Sheet — many rows of structured data you update regularly. The agent looks up the right row and answers from it.
  • The Catalog — a handful of curated products with images, meant for recommendations, not bulk lookups.
  • The FAQ — fixed questions with fixed answers that rarely change.
The sheet stays your source of truth. Change a price or mark a row sold in the spreadsheet, and the agent’s next answer reflects it — no re-training, no re-uploading. That’s the whole point of connecting a sheet instead of pasting the data in.

How it works

When you connect a sheet, your agent gets a search tool for it. The first row is read as column headers; every row below is a record. When a customer asks something the sheet can answer, the agent searches it and replies from the matching rows.

  • It’s a fuzzy search — the agent searches in plain language, in Hebrew or English. A customer doesn’t need to match your exact wording; “3 rooms with parking” finds the right rows.
  • The first columns matter most — earlier columns carry more weight in the search, so put what people search by (area, product name) on the left.
  • The agent only reads — when it answers a customer, it looks things up in your sheet; it doesn’t edit it. You can grant write access separately to sync data in — that’s the next section.
  • Updates land within a few minutes — the sheet is cached briefly for speed, so an edit shows up in answers after a short delay, not instantly.
The description is the brain: the short description you write when connecting is what tells the agent when to reach for the sheet and how to search it. It’s the highest-leverage part of the whole setup — there’s a section on it below.

Connect it

Before you connect, make the sheet readable. In Google Sheets, open Share (top-right), and under General access switch Restricted to “Anyone with the link,” with the role set to Viewer:

General access → “Anyone with the link,” role “Viewer.” This is Google’s own dialog.
Viewer is enough to read. That one setting is all the agent needs to search the sheet. The connect screen also shows an email to add as Editor — that’s the write permission, for syncing data into the sheet; it’s covered in the next section. For reads, “Anyone with the link → Viewer” is all you need.

Then connect it in the app:

Click "Build" at the top.
"Integrations" is in the side menu, under "Connect."
Paste the link, describe the sheet, and connect.

Pick Google Sheets from the integrations grid, paste your link, write a description (next section), and hit Connect Google Sheet. Done — the agent can now search it.

Let it write, too

Reading is one direction. You can also let data flow the other way — grant write access and we’ll keep the sheet in sync for you. It’s optional, and separate from how the agent answers.

In the same Google Share dialog, add the address the connect screen gave you and set its role to Editor. Leave General access on Viewer — you’re not opening the sheet up to everyone, just granting us edit rights:

Add our address as “Editor.” General access stays “Viewer.”

Back in the app, the connect screen confirms it with a green “Edit access is active.”

  • Viewer — the agent can read and search the sheet. Enough for lookups.
  • Editor — we can also write into the sheet and keep it current — for example, syncing your product catalog into it on a schedule, so the sheet always mirrors your live products.
The conversation never writes. Even with Editor access, the agent answering a customer still only reads — writing happens in a background sync you set up, never mid-chat. Your sheet is only ever changed by the sync you turned on.

A real example, end to end

Say you run a rental agency. Available apartments change every week, so they live in a spreadsheet your team keeps up to date. Here’s the whole thing, from sheet to answered customer.

1. The sheet

One row per apartment, headers on top. Notice the searchable columns — area, rooms, parking — sit on the left, and a Status column marks what’s still available.

One listing per row. You keep this updated; the agent reads it.

2. The description

When connecting, you tell the agent what this sheet is and when to use it. This one names the columns, says when to search, and how:

The link and the description the agent uses to search.

3. The live conversation

Now a customer messages on WhatsApp. The agent reads the intent, searches the sheet, and answers from the matching row — including whether it’s still available:

The agent searches the sheet and answers from the matching listings.
No dead data. When that 3-room gets rented, your team marks it Taken in the sheet — and within minutes the agent stops offering it. The conversation and the spreadsheet never drift apart.

Write a description that works

The description is the one field that decides whether this works well. It’s not decoration — the agent reads it to know when to search the sheet and how. Cover three things:

  • What each row is — “each row is one apartment, with area, rooms, price, parking, floor, and availability.”
  • When to use it — “whenever a customer asks what we have or filters by area, rooms, price, or parking.”
  • How to search it — “search by the area and features they mention.”
Weak vs strong: “Apartments.” tells the agent almost nothing. “Available apartments for rent — area, rooms, price, parking, floor, availability. Use it when a customer asks what we have; search by the area and features they mention.” tells it exactly when to reach for the sheet and what it’ll find.

You can edit the description any time from the connected view — if the agent misses a question it should have answered from the sheet, tighten the description first.

Test & keep it fresh

Before you rely on it, run through this:

  • Access — the sheet is “Anyone with the link → Viewer,” so the agent can actually open it.
  • Headers — the first row is real column names (not a title or a blank row), and the most-searched column is on the left.
  • A real question — ask the agent something only the sheet can answer, and check it pulls the right row.
  • A change — edit a cell, wait a couple of minutes, and confirm the new answer reflects it.
Keep it lean: the agent searches every column, so extra note columns and merged cells just add noise. A clean table with clear headers searches best.
Was this helpful?