Written by Marijn Overvest | Reviewed by Sjoerd Goedhart | Fact Checked by Ruud Emonds | Our editorial policy

Gemini in Google Sheets for Procurement: 8 Prompts That Save Hours a Week

Why Google Sheets is the Procurement Workhorse for Workspace Organisations

For procurement teams on Google Workspace, Google Sheets is where most of the analytical work happens. The spend extract from the ERP lands in Sheets. The supplier scorecard lives in Sheets. The contract KPI tracker is in Sheets. The category review analysis starts in Sheets. The RFP evaluation matrix is in Sheets.

Most procurement teams find that isolated experiments with Gemini only become a durable team capability when tool practice is paired with structured training. The AI Fundamentals for Procurement Teams program is built for exactly that transition, from individual curiosity to a procurement function that works differently.

The reason is the same for Google Workspace teams as for Microsoft 365 teams: spreadsheet work is the natural home for procurement analysis. Structured data, flexible analysis, collaborative editing. Gemini inside Google Sheets does for Workspace procurement teams what Copilot in Excel does for Microsoft 365 teams, it brings AI into the spreadsheet rather than requiring the procurement professional to move data out to an AI tool and back.

The Procurement Tactics 2026 AI Readiness in Procurement survey found that about 34% of procurement teams use Gemini today. Adoption is concentrated in Google Workspace-native organisations and growing fastest there. For procurement teams on Workspace that have not yet started using Gemini in Sheets seriously, it is usually the fastest path to measurable AI value in the procurement function.

The Eight Working Prompts

Eight prompts cover the bulk of procurement's Sheets work. Each follows a similar structure: a clear action, the target columns or range, and a defined output.

Prompt 1, Spend file cleaning

"Standardise supplier names in column B, merging entries that appear to be the same supplier with different spellings. Convert all amounts in column E to EUR using the exchange rate tab. Flag any rows where the supplier name or amount looks incorrect."

Typical output: a cleaned version of the source file plus a review column flagging uncertain entries. The procurement analyst validates the flagged entries before proceeding with analysis. Time saving on a 10,000-line spend file: a day of manual work compressed into under an hour including validation.

Prompt 2, ABC-XYZ classification

"Classify each SKU in this file using ABC-XYZ methodology. ABC should rank by annual spend contribution: A = top 80% of spend, B = next 15%, C = remaining 5%. XYZ should rank by demand predictability across the 24 months in columns F through AC: X = low coefficient of variation, Y = medium, Z = high. Add two columns with the classifications."

Output: classified file ready for category analysis. Standard procurement classification work compressed from a pivot-and-formula exercise to a single prompt.

Prompt 3, Supplier concentration analysis

"Produce a summary of supplier concentration for this spend file. Show: top 10 suppliers as share of total spend, top 50 as share, concentration by category (top 3 suppliers in each of the six top-level categories), and count of single-source suppliers. Output as a table on a new sheet."

Output: the concentration view that anchors every category review. Consistent across analysts, consistent across months.

Prompt 4, Contract KPI scorecard

"Build a scorecard comparing contractual KPIs (from the 'Contract Terms' tab) against actual performance (from the 'Actuals' tab) for each supplier in the 'Active Suppliers' tab. For each KPI, show: contractual target, 12-month actual, variance, and status (green/amber/red based on the tolerance rules in the 'Definitions' tab)."

Output: scorecard ready for supplier review or QBR preparation. The work that used to take an analyst half a day happens in minutes.

Prompt 5, Savings tracking

"Calculate realised savings against baseline for each category in this spend file. Baseline is in column D; actual spend is in column E. Show: total savings, savings rate, and savings vs target by category. Flag categories where actual savings fall below 50% of target."

Output: the savings view that feeds finance reporting. Consistent methodology, repeatable across periods.

Prompt 6, RFP scoring matrix auto-fill

"For each supplier response in the 'RFP Responses' tab, apply the scoring framework in the 'Evaluation Framework' tab and populate the 'Scoring' tab with dimension-level scores plus commentary. Flag any responses where specific required information is missing."

Output: first-pass RFP scoring that the evaluation team validates rather than produces. The meeting starts with the scoring in place; the discussion focuses on the judgement calls.

Prompt 7, Tail-spend detection

"Identify suppliers in this spend file representing less than 0.1% of total spend individually. For each tail supplier, show: category, annual spend, whether a preferred supplier exists for the same category. Output as a tail-spend report sorted by consolidation potential."

Output: the tail-spend view that feeds consolidation programmes. The work that procurement analysts know they should do quarterly and rarely have time for.

Prompt 8, Spend forecasting

"Forecast spend for the next four quarters for each of the six top-level categories using the 24 months of historical data in columns F through AC. Account for seasonal patterns and the growth trend. Output as a table with quarterly forecast values plus a confidence indicator."

Output: a defensible spend forecast for budgeting conversations. Not a replacement for human judgement about known category changes, but a credible baseline to adjust from.

Getting Gemini to Produce Useful Output Reliably

Three practices separate procurement analysts who get consistent results from Gemini in Sheets from those who do not.

Cleanup before the run. Gemini produces clean output from clean data. Standardised supplier names, single currency, consistent category taxonomy, defined time period. The ten minutes of cleanup before the prompt is typically the highest-leverage ten minutes of the whole exercise.

Specificity in the prompt. Generic prompts produce generic output. Prompts that specify the target columns, the expected output format, and the calculation rules produce reliable output. "Classify this spend file" produces mixed results; the prompt in the ABC-XYZ example above produces consistent results.

Validation after the run. Spot-check the top-line numbers against pivots on the source data. If the totals reconcile, the detailed output is usually reliable. If they do not, the source data has a structural issue that needs to be fixed before the analysis is used. Procurement analysts who skip validation produce reports that get challenged in the first review and erode credibility.

Where Gemini in Sheets Wins Against Alternatives

Against Claude or ChatGPT, Gemini in Sheets has a clear workflow advantage: the analysis stays in the spreadsheet. No copying outputs between tools. No version-control question about which copy of the file is authoritative. For procurement analysts working primarily in Sheets, this is decisive.

Against Copilot in Excel, the choice is largely determined by productivity suite. Microsoft 365 teams naturally use Copilot; Google Workspace teams naturally use Gemini. The quality of the underlying analysis is broadly comparable.

Gemini's one distinctive strength in this comparison is long-context handling. For very large spend files, hundreds of thousands of rows, Gemini's context window handles more of the data in a single pass than some alternatives. For most procurement work this does not matter; for procurement functions with detailed transactional data, it can.

Where Gemini in Sheets Falls Short

Three honest limitations.

Gemini sometimes makes arithmetic errors that look right. A concentration percentage computed against the wrong denominator, a rolling average that includes the wrong time window, a classification threshold misapplied. The errors are subtle, they produce output that looks plausible, which is why validation against the source data is essential.

Category-specific expertise is not Gemini's strength. Gemini applies generic procurement analysis patterns. A category-specific nuance, commodity-market-driven pricing logic for metals, contract structures specific to IT services, healthcare-specific supplier compliance frameworks, is not inside Gemini's default knowledge. The procurement analyst brings that; Gemini provides the mechanical analysis on top.

Very complex multi-step analyses sometimes lose coherence. If the prompt asks Gemini to do five distinct things in sequence, the output sometimes misses steps or applies them inconsistently. Breaking complex analyses into two or three prompts produces more reliable output than trying to do everything in one.

None of these limitations undermines the value proposition. They are constraints on how procurement analysts should use the tool, not reasons to avoid it.

Frequently asked questions

How large a spend file can Gemini in Sheets handle?

Performance is good up to roughly 100,000 rows in a standard file. Above that, splitting the analysis by category, time period, or business unit produces more reliable results. Gemini's long-context capability helps but does not eliminate the practical limits.

Is Gemini in Sheets appropriate for sensitive procurement data?

On an enterprise Workspace plan with appropriate data-handling terms, yes, for most categories of procurement data. The same commercial data-handling terms that cover Gmail and Docs cover Sheets. Highly sensitive commercial data may still warrant separate policy review.

How accurate is Gemini's spend classification?

For most spend files with reasonable supplier and category data, classification accuracy is high enough that validation of flagged entries is a small share of the work. For spend files with unusual taxonomies or low-quality supplier descriptions, more validation is needed, but the starting point is still faster than fully manual classification.