When Gemini in Sheets speaks Swedish: turn customer rows into a work list

It sounds like a minor language update. It is not.

Google says Fill with Gemini and the AI function in Google Sheets are now rolling out to eleven more languages, including Swedish, Danish, and Norwegian. For Nordic teams, that matters. A normal customer spreadsheet can now become more than a graveyard of rows: it can become a first draft of work, in the same language customers actually use.

This is not a call to let AI "take over customer service." That is the wrong starting point. The better start is more practical: take twenty or fifty customer rows, let Gemini summarize, categorize, and suggest a next step, then have a human review the result before anything is sent, changed, or promised.

Source: Google Workspace Updates: Fill with Gemini in Sheets now available in 11 additional languages

What Google actually changed in Sheets

Google describes two capabilities that now become more useful for Nordic teams. Fill with Gemini can help users fill columns based on the context of a table. The AI function in Sheets can generate text, summarize information, categorize inputs, analyze sentiment, and access current information from Google Search.

These features were previously available in English, Spanish, Portuguese, Japanese, Korean, French, Italian, and German. On July 7, 2026, Google announced that both Fill with Gemini and the AI function are expanding to Mandarin, Dutch, Malay, Hebrew, Polish, Turkish, Czech, Indonesian, Swedish, Danish, and Norwegian. The gradual rollout started on July 7 and can take up to 15 days.

The main point for Hammer readers is not the language list by itself. It is that AI in spreadsheets becomes more useful in the everyday language where the work already sits: customer emails, quote notes, form responses, student feedback, booking notes, complaints, and follow-ups.

Source: Google Docs Editors Help: Use the AI function in Google Sheets

Start with customer rows, not a grand AI plan

Many organizations already have a basic customer list in Sheets. It may come from a form, a CRM export, an inbox someone pasted from, or a manual list after an event. The usual problem is not that the list does not exist. The bottleneck is that it does not turn into work.

One row might say:

"Wants pricing for a service contract, prefers calls after 3 p.m., mentioned the previous invoice was wrong."

Another might say:

"Parent asks whether the course material can be provided in English before next week."

That is not just data. It is a set of small decisions waiting for someone: who should answer, what should be said, what is urgent, what needs to be checked in another system?

Gemini in Sheets fits this problem because each row is already a bounded piece of work. You do not need to build an agent platform on day one. You need a better first sort.

A 45-minute test: from rows to work list

Run this in a copy of a spreadsheet, not in the original. Choose one focused workflow where the result can be reviewed before anything moves on. Good candidates are customer questions from a website form, course evaluations, support cases, quote requests, or store comments.

1. Create a copy and add four new columns

Add Summary, Category, Next step, and Reviewer. If you want to track quality, also add Accepted, Changed, and Comment.

2. Define the categories before AI gets involved

Choose five to seven categories. For a customer list, that might be: reply today, follow up later, quote, invoice question, product idea, complaint, and unclear. The point is to stop the AI from inventing a new category for every row.

3. Use the AI function in test columns

Google shows the syntax as =AI("prompt", [optional range]) or =Gemini("prompt", [optional range]). Some local spreadsheet settings may require semicolons instead of commas, so adjust the separator to your environment.

For example, paste this into a test column and replace A2:D2 with the cells that contain the customer text and relevant context:

=AI("Summarize the customer's issue in Swedish in one sentence. Keep product names, dates, and amounts if present. If the source is unclear, write 'unclear source'.", A2:D2)

For the category:

=AI("Classify the case as exactly one of these categories: reply today, follow up later, quote, invoice question, product idea, complaint, unclear. Reply only with the category name.", A2:D2)

For the next step:

=AI("Suggest the next step in Swedish. Max 18 words. Do not promise anything that is not in the source. If a human needs to check something first, state what should be checked.", A2:D2)

4. Review twenty rows manually

Read row by row. Mark whether the AI suggestion was useful, needed edits, or was wrong. Add a short comment when it failed. This is not bureaucracy. It is how you discover which cases AI actually helps with.

5. Build the work list, not the automation

Sort by category and next step. Add a responsible person in Reviewer. Only when you see that one category works repeatedly should you consider the next integration: create a task, update the CRM, send an internal notification, or prepare a customer reply.

Copy the prompt: turn the Sheets list into a reviewed work queue

Paste this into Gemini or your usual AI tool before you build the formulas. Ask it to adapt the formulas to your actual columns.

You are my workflow designer for Google Sheets.

The goal is to turn a customer or case list into a reviewed work queue, not to automatically reply to customers.

These are the columns in the spreadsheet:
[paste the column names]

Here are 5 example rows, anonymized if needed:
[paste examples]

Suggest:
1. Four new columns that make the list easier to work with.
2. One AI formula per column with a clear Swedish instruction.
3. A fixed category list with no more than seven categories.
4. A simple review routine for the first 20 rows.
5. Which results should never be sent externally without human approval.

Be practical. No strategy fluff. I want to build the test in a copied spreadsheet in 45 minutes.

Where the boundary should be

Google notes a few practical limits in its help documentation. The AI function returns text. It does not have access to your entire spreadsheet or other Google Drive files unless you provide the relevant data in the sheet and in the function argument. You cannot expect it to understand context that only exists in someone's head.

That is useful. It forces better input. If a row does not include the customer's question, product, date, or status, the AI should write "unclear source" instead of guessing. A good routine makes gaps visible.

For integrations, one rule is enough: AI may prepare work, but a human approves external replies, price changes, bookings, invoice adjustments, and updates in systems where mistakes become expensive. When you connect the workflow to CRM or automation, use separate accounts, scoped permissions, environment variables or secret managers for keys, and a run log that shows who approved what.

That is not fear. It is good operations.

Three rows worth testing first

Customer questions from forms

Let AI summarize the case, set a category, and suggest the next internal step. A human writes or approves the reply.

Course or school feedback

Let AI group comments into themes: unclear material, pace, technical problems, requests, and praise. Use the result to plan the next lesson or message.

Quote and service requests

Let AI mark which information is missing: address, time, budget, current system, invoice number, or decision date. Follow-up becomes shorter and more concrete.

What this means for Hammer readers

The useful part of this update is that it does not require a completely new way of working. Many teams already have Sheets. Many already have rows waiting. The difference is that Swedish, Danish, and Norwegian now sit closer to the place where everyday work happens.

If you want to use AI more practically in the business, this is a good first week: choose one list, create a copy, add four columns, review twenty rows, and decide what is worth repeating. If the result holds up, the next step can be a lightweight Tool Forge setup: the same reviewed work queue, connected to tasks, CRM, or internal notifications with clear approvals.

Start where the rows already are. That is often where automation becomes real.