Spreadsheet Productivity: Sheets Cites, Excel Cleans

On October 28, 2026, Google announced cell-level source citations and an expanded 3,000-cell capacity for Fill with Gemini in Google Sheets. On October 27, 2026, Microsoft published its monthly feature release bringing the Clean Data skill in Microsoft Excel to General Availability across Windows, Mac, and the web. Both releases focus directly on spreadsheet productivity, addressing an operational problem in daily workbook maintenance: unverified text outputs and structural formatting errors often create more cleanup work than automated generation saves.

When spreadsheets ingest automated entries without clear verification trails, office workers spend hours double-checking calculations. As detailed in our earlier look at how AI tools free up time only when workers avoid manual reconstruction, raw speed provides little benefit if the output requires line-by-line validation. The updates from Google and Microsoft approach this friction from opposite sides of the workbook. Google aims to make external information verifiable by embedding source links directly inside individual cells, while Microsoft targets the hidden whitespaces, duplicate rows, and conflicting date formats that break lookup formulas.

The October gap between deployment and daily spreadsheet productivity

The timing of these product updates matches findings from workplace research published in late October. On October 21, 2026, Ricoh USA released findings from its 2026 Workplace Technology Adoption Gap survey, exploring the divide between executive expectations and daily worker experiences across North America.

The survey gathered responses from 2,017 adults employed full-time at organizations with at least 10 employees across the United States and Canada in September 2026. The participant pool included 1,511 employees (1,005 in the U.S. and 506 in Canada) consisting of individual contributors, non-manager employees, and team leads below the director level who use workplace technology in their daily roles. It also included 506 business and IT decision-makers (375 in the U.S. and 131 in Canada) at the director level and above who hold decision-making influence over AI, workplace tools, operations, or digital transformation.

According to the Ricoh survey, 91% of business decision-makers expressed confidence that their investments in AI tools deliver the value they expected. In contrast, 53% of employees reported that AI creates friction or extra work. Furthermore, 55% of employees stated that time spent managing workplace technology tools actively interferes with their productivity. Among the four workplace technology categories evaluated in the study, which also included communication and collaboration, information access, and workflow management, AI tools delivered the lowest employee-reported productivity gains.

System disconnects contribute heavily to this day-to-day friction. Three in five employees (60%) reported that workflow and process-management tools create friction or extra work. In addition, 56% of employees stated that information, knowledge, and data access tools create friction or extra work in their daily tasks. Leadership and staff viewed system integration differently: 90% of decision-makers believe their organization’s technology tools, systems, and processes are connected, whereas 76% of employees stated that their jobs would be easier if those same tools, systems, and processes worked better together.

When official workplace systems prove cumbersome or disconnected, employees frequently seek workarounds. The Ricoh report found that 45% of employees use non-approved tools to get work done more easily at least monthly. This gap extends to governance and safety guidance. While 90% of decision-makers believed their staff have clear guidance on how to use AI-enabled tools responsibly, only 66% of employees agreed.

The underlying organizational friction stems from how technology rollouts are managed. While 71% of decision-makers stated that their organizations usually seek employee input before rolling out new technology, only 43% of employees reported that this consultation usually occurs. Moreover, 81% of decision-makers acknowledged that their organization invests in new workplace technology faster than it helps employees learn to use it effectively.

When workers receive automated software without structured verification processes, spreadsheet files quickly become unstable. A finance manager who imports hundreds of automated supplier entries must spend entire afternoons cross-referencing values against primary records. This repetitive checking adds to the hidden productivity cost of context switching, pulling analysts out of deep work to resolve basic database discrepancies. The late October updates from Google and Microsoft target these practical problems by introducing cell-level citation links and automated data-cleaning routines.

Google Sheets attaches citations directly to 3,000 cells

Google’s update, published on the Google Workspace Updates blog on October 28, 2026, expands the Fill with Gemini feature in Google Sheets. The feature uses natural language prompts to categorize information, summarize records, and populate tabular data.

The October 28 release changes how the tool handles workbook requests in two practical ways.

First, when users employ the AI function to generate cell values, Gemini provides the exact web sources referenced directly within each individual cell. By linking citations directly to individual cells, the web interface gives users a straightforward way to inspect sources, cross-check figures against original pages, and verify data before sharing workbooks.

The update also expands processing capacity to support up to 3,000 cells per request, allowing users to summarize, categorize, and process larger datasets in a single prompt. Google reported that the upgraded backend returns answers faster, maintains greater consistency across adjacent cells, and completes 500-cell batches up to 55 times quicker than manual data entry. According to Google’s technical notes, this benchmark is based on an internal study of 95 participants comparing manual entry against Fill with Gemini on a 500-cell task using English prompts for English language users.

Users can access the tool in the side panel by clicking on “Ask Gemini” (the spark button) in the top right corner of Google Sheets. Gradual rollout started on October 20, 2026, with an expected timeline of up to 15 days for feature visibility across both Rapid Release and Scheduled Release domains. The update is available to Business Standard, Business Plus, Enterprise Standard, Enterprise Plus, Google AI Pro, Google AI Ultra, Google AI Pro for Education, and accounts with the AI Expanded Access add-on.

The operational value of cell-level attribution lies in immediate auditing. In earlier iterations of workspace assistants, such as those analyzed when Gemini began providing automated daily summaries, outputs appeared as unlinked text strings. If an assistant inserted an outdated vendor fee or incorrect corporate address into row 42, an analyst had to open an external search engine to confirm where the figure originated. With cell-level attribution, verification occurs directly inside the web spreadsheet interface.

Excel scans spaces, types, and dates before formulas break

While Google focused on external information retrieval and citation, Microsoft concentrated on internal workbook data hygiene. On October 27, 2026, the Microsoft Tech Community Excel Blog published its What’s New in Excel (October 2026) release, announcing that the Clean Data skill for Copilot is rolling out to General Availability across Excel for Windows, Mac, and the web.

The Clean Data skill inspects relevant data ranges for five specific structural issues:

Extra spaces that cause exact-match lookup functions like VLOOKUP or XLOOKUP to return #N/A errors.

Inconsistent text and number formats, where numeric quantities are stored as text strings alongside raw numerical values, corrupting calculations such as SUM or AVERAGE.

Inconsistent dates, where mixed serial formats prevent chronological sorting and produce miscalculated intervals in financial modeling.

Duplicate values that inflate record totals in transaction databases and financial reconciliations.

Values stored as the wrong data type, such as numeric codes stored under incompatible structural definitions.

To invoke the tool, users enter @clean-data into their prompt to Copilot or select the skill from the built-in Skills menu. According to Microsoft’s release notes, the skill highlights potential errors, outlines how those issues could disrupt formulas, and suggests fixes directly in the sheet. Users review the proposed fixes and select which changes to apply, avoiding unintended workbook modifications.

Alongside the data-cleaning tool, Microsoft added a structural convenience for desktop users. Moving out of preview, Excel for Windows now lets users name tables immediately upon creation. When creating a table through Ctrl+T or the Insert > Table command, users can enter a specific name right away. Naming a dataset “Regional_Sales_2026” at setup prevents the workbook from assigning generic default labels like “Table1,” making structured formula references readable across large financial sheets.

Skill recorder and custom routines replace repetitive prompt typing

The October 27 Excel release also addresses repetitive prompt entry. When knowledge workers use conversational sidebars to reformat weekly reports, typing the same instructions every Monday morning consumes valuable time. To automate these sequences, Microsoft detailed several custom workflow tools in its October release and Microsoft Support documentation for Copilot skills.

With Skill Recorder, users can show Copilot how they handle a recurring task. As the user carries out the task, Copilot tracks each action and converts the sequence into an automated skill. People can either demonstrate the procedure on-screen or summarize the steps in writing, letting Copilot repeat the routine reliably on future workbooks. Skill Recorder is currently available to Insiders of Excel for Windows and Excel for Mac, and is generally available on Excel for the web.

Complementing the recorder, Skill Builder allows users to tell Copilot what they want done. Users can submit a preferred workflow, written instructions, past prompt queries, or finished sample sheets, and Copilot turns that guidance into an automated routine. Skill Builder is currently rolling out to general availability across all platforms and can be called using @skill-builder.

According to Microsoft Support guidelines, custom skills follow a standardized structure. Skills are saved to the user’s OneDrive account in a dedicated skills folder. Each skill requires an individual folder containing a SKILL.md file. The file includes a required frontmatter metadata block between three dash delimiters (---) containing the name and a description that informs Copilot when to trigger the skill. The body section beneath the frontmatter provides the specific step-by-step instructions telling Copilot how to execute the task.

For data import workflows, the update includes the Get Data skill on Excel for the web. Users describe the data they need and how they want it prepared, and the skill connects to the source, transforms the dataset, and loads the results into the workbook using Power Query. This creates an inspectable query that users can edit and refresh directly without manually assembling connection strings.

Microsoft also noted that it is expanding model choices by bringing GPT-6.1 Sol and Claude Sonnet 5.5 to Microsoft Copilot, providing additional options for delegating work to Copilot. For visual summaries, Excel for Windows Insiders received Interactive Canvas, which transforms workbook numbers into dynamic presentation views that remain tied to the underlying sheet and refresh automatically whenever values change.

Comparing the verification models: Inline auditing versus structural cleaning

The late October updates take different directions to improve workbook reliability. Google treats the spreadsheet as an information retrieval canvas that imports external facts into rows, whereas Microsoft treats the spreadsheet as a calculation engine that requires clean inputs to keep formulas functional.

Feature Dimension	Google Sheets (October 28, 2026)	Microsoft Excel (October 27, 2026)
Primary Focus	Web source citation and batch data generation	Structural data cleaning and formula error prevention
Core Feature	Fill with Gemini	Clean Data Skill (@clean-data)
Processing Limit	Up to 3,000 cells per request	Selected workbook ranges and target tables
Performance Benchmark	Up to 55x faster than manual entry for 500 cells	Automated detection across five data integrity categories
Output Verification	Exact web URLs attributed to individual cells	Proposed workbook corrections for review and selective application
Automation Building	Prompt-based batch processing in side panel	Skill Recorder, Skill Builder, and Power Query Get Data
Storage and Standards	Integrated directly within Google Workspace	Custom skills saved as SKILL.md in OneDrive
Platform Availability	Google Sheets on the web across eligible plans	Windows, Mac, and web across Copilot subscriptions
Rollout Schedule	15-day gradual rollout starting October 20, 2026	General Availability rollout announced October 27, 2026

Google’s cell-level citation model addresses unverified data claims. When a team uses Fill with Gemini to assemble a market comparison table covering 150 software products, each cell records the exact URL used to generate that entry. If a cell displays a subscription cost or corporate headcount, any colleague can click the embedded web link to verify whether the number came from an official pricing page.

Microsoft’s clean data model addresses structural workbook corruption. When team members import transaction records from external databases, trailing whitespaces and mixed calendar formats regularly disable formulas. Running @clean-data cleans text strings and standardizes dates before those inconsistencies produce calculation errors across shared reports.

Practical audit rules for daily spreadsheet workflows

Applying these spreadsheet updates effectively requires consistent workplace habits. Based on the survey findings and product mechanics released in October 2026, teams can maintain workbook integrity by establishing four specific operational routines:

1. Verify cell citations before sharing external research sheets

When using Google Sheets to compile research across large cell ranges, treat AI generation as a preliminary step. Click into generated cells to inspect the attached URL citation. Check that the source link leads to an official company portal, regulatory register, or verified publication before passing the workbook to clients or department heads.

2. Run clean data checks immediately after importing raw data

Do not build PivotTables or multi-sheet lookup formulas on raw CSV imports before checking data hygiene. In Microsoft Excel, invoke @clean-data directly after pasting new transaction logs. Review the suggested fixes for trailing spaces and mismatched dates, and apply the corrections before writing downstream lookup formulas.

3. Convert recurring data routines into custom skills

If your team spends fifteen minutes every week reformatting transaction tables, save the steps as a reusable skill. Use Excel’s Skill Recorder to capture sorting, filtering, and column cleanup steps, or create a SKILL.md file in your OneDrive skills folder with clear frontmatter instructions. Running a saved custom skill ensures consistent formatting without retyping prompts each week.

4. Assign descriptive table names during creation

Avoid leaving datasets with default names like “Table1” or “Table2.” In Excel for Windows, use the creation prompt to enter a descriptive name such as “Q4_Regional_Expenses.” Descriptive table naming makes structured formula references clear and simplifies auditing across shared department drives.