"When I asked in Japanese to 'aggregate last month's sales by region into a table,' Gemini actually built the spreadsheet for me. Does this mean we won't need contractors or in-house Excel experts anymore?"—an executive at a client company recently asked me this. I understand the feeling well. Until now, spreadsheet summary tables were a "sanctuary" accessible only to a few individuals who could write VLOOKUP, QUERY, or macros. Now, tables, pivot tables, and charts can all be constructed using natural Japanese instructions. The barrier to entry has indeed dropped dramatically. However, right then and there, I told him that being able to create something and being able to rely on it in business operations are two different matters. Rather, from the perspective of those managing operational sites, our frank impression is that a new kind of tacit reliance is about to begin.
On June 18, 2026, Gemini in Google Sheets added support for creating and editing entire spreadsheets in 27 languages including Japanese (totaling 28 languages alongside existing English). As of April of this year, construction and editing capabilities such as creating tables and pivot tables, generating charts, and inserting formulas were already available in English, but this update makes them operable via natural Japanese prompts (the gihyo.jp article covers the details clearly). In other words, the process of building spreadsheets, once reserved for those who could write formulas, has turned into a task you can delegate with words. That is precisely why the role of custom development is shifting away from building tables on a client's behalf to something else. In this article, we outline this shift and what is required to use Gemini-created spreadsheets in business with confidence.
What Has Changed, and What Remains Unchanged
First, let us pin down accurately what is now possible. When you prompt Gemini in the side panel in Japanese, Gemini does not immediately start building; instead, it returns suggestions to refine your prompt first. For example, it might ask, "For this aggregation, would it be easier to view if broken down by month? How would you like to proceed?" Once you reply, it then presents a work plan, and only after you approve does it create and edit the sheet according to that plan. Having a workflow that inserts a plan rather than immediately rewriting your sheet is a thoughtful design.
There are a few prerequisites for use. Smart features must be enabled in Workspace; only when turned on does the Gemini icon appear in the side panel. In addition, higher usage limits are set as a trial period until July 15, 2026, after which per-user usage limits will apply. Note that Gemini itself has been standardly integrated into each Google Workspace plan since January 2025, eliminating the former need to purchase add-ons separately.
So what remains unchanged? The fact that someone must guarantee whether the contents of the spreadsheet are correct has not changed at all. Rather, this is where the real issue begins. Previously, the creator and the person who understood the logic were the same person, as the individual who could write formulas built it and understood its inner workings. When Gemini can build spreadsheets, this alignment collapses. You can easily end up in a situation where the table exists, but no one understands what its formulas are calculating or under what assumptions the data is aggregated.
The New Tacit Reliance Surrounding Gemini-Created Spreadsheets
For years, the problem with spreadsheet operations was the siloed, person-dependent summary table that only staff who understood VLOOKUP or QUERY could touch. The moment that person transferred or resigned, no one could fix the content. This is a pain point most small and mid-sized businesses have experienced at least once.
With Gemini able to build tables, this person-dependency appears to be resolved. In reality, however, the form of dependency has merely shifted. Now, a new tacit reliance emerges where no one verifies the content of the table Gemini produces. Because not even the person who prompted it can read the formulas, it is in some ways more troublesome than before. While the previous dependency meant there was at least one person who understood it, the new dependency risks creating a state where not a single person understands it.
Where, specifically, do accidents happen? Here are the scenarios I watch out for in operational settings.
One case is where a formula generated by Gemini looks correct at first glance but fails under edge cases. For instance, it might handle columns mixing tax-exclusive and tax-inclusive numbers poorly, or omit rows from aggregations when blank cells or unexpected strings are present. While it may look correct on a sample of a dozen rows, the breakdown only surfaces when dealing with thousands of production rows. Another case is when users bypass the approval flow or approve without checking thoroughly, letting Gemini directly rewrite the production summary sheet and breaking original formulas. Even more insidious is exceeding usage limits. An automated process heavily relied upon for month-end closing hits the usage limit after the trial period ends and abruptly stops, paralyzing operations. The more you depend on convenience, the greater the fallout when it halts.
Furthermore, data handling is frequently overlooked. Teams start using it without determining whether they should allow Gemini to process sheets containing confidential transaction data or personal information with smart features enabled, or what data is referenced and processed. Jumping in simply because it is convenient leaves you unable to explain your practices during client audits later.
The Role of Custom Development Shifts from Building to Verification and Scoping
Here, the role of custom development changes. On-site staff can have Gemini build the tables themselves. Where we deliver value is before and after that—namely, on both sides flanking the act of creation.
One area is accuracy verification of the formulas and aggregations produced by Gemini. Instead of taking output tables at face value, we confirm whether they behave as intended under edge cases (blank cells, unexpected values, overflows, tax handling) and verify that they will not break under production data volumes. This work can only be performed by someone who can read formulas and knows empirically where things break down. For interpreting the formulas themselves or deciding when to supplement Gemini's shortcomings with GAS, the formula and scripting fundamentals covered in our Google Sheets automation article serve as the foundation for an evaluative eye.
Another area is demarcating smart feature toggles and data handling. Which sheets (and which data) can Gemini handle, and which must it not touch? We establish policies for sheets containing confidential or personal information in advance as clear rules so teams do not hesitate. The broader view of how to use Workspace AI safely connects with the approach outlined in our Workspace Gemini utilization guide.
Going a step further, auditing and standardizing existing, person-dependent sheets also falls under custom development. Introducing Gemini into an environment where numerous spreadsheets exist whose inner workings no one can explain only increases the verification burden. We first audit existing sheets, clarify what each table is for and what each formula calculates, and then separate what can be reproduced or replaced with Gemini from what cannot. This steady groundwork is where external expertise brings real value.
Discerning Where Gemini Suffices Versus Where Bespoke Development Is Needed
In custom development, the most effective contribution may actually be scoping what to delegate to Gemini and what to build using other means. Just because Gemini can build spreadsheets does not mean completing every operational task on Google Sheets is the right answer.
A mid-sized manufacturing company (name withheld) came to us with this dilemma: "Frontline staff started using Gemini to create inventory summary tables one after another, but tables proliferated so much that no one knows which is the correct, latest version. Spreadsheets that should show the same inventory numbers report different figures depending on the sheet." Lowering the barrier to building tables with Gemini led to a flood of superficial variations, creating confusion instead.
What we did was not take Gemini away. First, we audited the proliferating inventory sheets and consolidated them to establish what the single source of truth was. Then, for ledger-type operations where multiple people input data and state continuously updates—such as daily inventory check-in/out records—we shifted away from Google Sheets plus Gemini toward dedicated applications, similar to those covered in our article on operational apps with AppSheet. On the other hand, for one-off aggregations and visualizations, such as aggregating monthly sales by region and product to prepare materials for executive meetings, we organized procedures so staff could build them in Japanese, as this is where Gemini excels.
Why does this scoping work? Because Gemini is strong at ad-hoc aggregation and visualization, but unsuited for operations where multiple people continuously enter data and data integrity must be maintained. Forcing Gemini to keep building the latter within spreadsheets deepens confusion as tables multiply, as seen in the inventory example. Move ledger tasks to dedicated apps, and leave aggregation and visualization to Gemini. Drawing this line at the beginning preserves data integrity while leveraging Gemini's convenience. For maintaining legacy automations built with GAS, please also refer to our article on maintaining inherited Google Apps Script.
What to Verify Before Considering Adoption
If you are looking to streamline spreadsheet operations with Gemini going forward, checking just two things before you start building will save you from painful lessons later.
One is to determine who will verify and guarantee the contents of the tables Gemini creates. Putting sheets into production merely because they could be created leads to a proliferation of tables no one can read, ultimately relapsing into a new form of person-dependency. Incorporating verification under edge cases and production data volumes as part of your operational workflow is crucial. The other is to demarcate the scope of data permitted to be handled with smart features enabled beforehand so teams do not hesitate. Policies for sheets containing confidential or personal information must be decided before getting swept up in convenience.
Whether Gemini can now create tables but no one internally can verify their content, Gemini-created summary tables have proliferated to the point where the latest version is unknown, or you want advice on scoping ledger operations versus ad-hoc aggregations, please contact us through GleamHub's inquiry page. We will review your current spreadsheets and operational workflows, collaborate to sort out what can be entrusted to Gemini, what humans must verify, and what should be shifted to other tools, and help establish a setup your teams can continue using with confidence.
