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

Copilot for Spend Analysis in Excel: The Procurement Playbook

Where Spend Analysis Hours Actually Go

Ask a procurement analyst what takes the most time in a typical month and the answer is usually some version of spend work. Consolidating ERP extracts with supplier master data. Classifying line items into procurement categories. Normalising supplier names when "Acme Corp", "Acme Corporation", and "ACME Ltd" all appear in the same file. Building the concentration view that shows which suppliers and which categories carry the most spend. Producing the tail-spend report that explains why a third of the spend is sitting with two hundred small suppliers.

None of that work is intellectually difficult for a procurement professional. It is high-volume, pattern-based work that rewards attention to detail and consistency. It is also exactly the kind of work Copilot in Excel handles well, because Copilot sees the spreadsheet, has full access to the data in it, and can apply pattern-based transformations at scale.

The Procurement Tactics 2026 AI Readiness in Procurement survey found 40% of procurement teams spend 60% or more of their week on manual data work and reactive firefighting. A substantial share of that time is spend work. In procurement teams with a dedicated analyst function, it is most of the analyst's week. Copilot compresses the pattern-based majority of spend work and leaves the analyst with time for the strategic layer on top, hypothesis generation, savings identification, and category planning.

What Copilot in Excel Actually Does for Spend Analysis

Five categories of spend work are where Copilot in Excel earns its place fastest.

ABC-XYZ classification

Standard procurement classification: ABC ranks SKUs or categories by volume contribution to spend; XYZ ranks by predictability of demand. Copilot runs both in a single prompt, adding the classification as new columns in the working sheet. For a typical ten-thousand-row spend file, this is seconds of Copilot time, the procurement analyst would take hours to do it with pivots and formulas.

Supplier concentration analysis

The 80/20 view of the supplier base, top ten suppliers as a percentage of total spend, top fifty as a percentage, concentration by category, single-source versus dual-source share. Copilot produces the analysis as a summary table; combined with a brief prompt to chart the output, it produces the concentration visual too.

Tail-spend detection

The suppliers below a threshold, commonly suppliers representing less than 0.1% of total spend individually, that in aggregate represent a significant portion of the supplier base and a smaller but non-trivial share of spend. Tail-spend analysis has a well-known shape but requires tedious threshold application. Copilot runs it as a single prompt and produces the tail-supplier list with recommended consolidation candidates.

Maverick spend identification

Off-contract purchases, duplicate supplier usage for the same category, purchases split below an approval threshold, and purchases from suppliers not on the preferred supplier list. Each is a distinct pattern, and Copilot can scan the spend file for all four in a single multi-part prompt. The output is a list of spend lines flagged for review.

Savings hypothesis generation

The output of the four analyses above is a set of opportunities, consolidation candidates from tail spend, maverick spend to recapture, high-concentration categories where a second supplier would create leverage. Copilot produces the opportunity inventory and a rough savings estimate for each, which the procurement analyst then validates against their market knowledge.

Each of these five is a single Copilot prompt. A procurement analyst running all five against a cleaned spend file is producing a first-pass spend analysis in roughly an hour, which is before any of the analysis work traditionally considered "spend analysis" would have started in the manual version.

The Copilot-Ready Spend File

Copilot runs well on clean data and struggles on dirty data. The ten-minute cleanup before Copilot runs is usually the highest-leverage ten minutes of the whole exercise.

Consistent supplier names. A supplier master data join is the ideal solution; a Copilot prompt to standardise supplier name variations is the practical fallback when the master data is not clean. The latter produces a "suggested standardisation" column that the procurement analyst reviews before applying.

Single currency. Multi-currency spend files cause Copilot to produce arithmetic errors because it defaults to summing mixed currencies as if they were the same. Convert to the reporting currency before the analysis runs.

Consistent category taxonomy. If the spend file uses inconsistent category labels (Packaging, packaging, PACKAGING, Pack-Ing Materials), Copilot spends effort reconciling them before it can run the actual analysis. A category taxonomy lookup table, applied before the Copilot run, cuts this.

Time period defined. Twelve months rolling is the usual default. A time period that includes partial months or overlaps fiscal year boundaries produces analysis artefacts that are hard to interpret.

None of these prerequisites are heavy. They are disciplines that any procurement analyst running spend analysis regularly already applies. Copilot's performance is simply more sensitive to them than a human analyst's, a human adapts to dirty data; Copilot produces dirty output when given dirty data.

What the Finished Spend Analysis Looks Like

An anonymised example illustrates the shape of the output.

A procurement organisation with roughly €180 million of annual spend runs a monthly spend refresh. The clean spend file, loaded into Excel with Copilot, produces the following in under an hour:

The ABC-XYZ classification shows 18% of SKUs driving 68% of spend (A-category), with the AX intersection, high-spend, high-predictability, accounting for 45% of total spend. This is the core of the spend base where supplier consolidation and strategic contracting should focus.

The concentration analysis shows the top ten suppliers account for 54% of spend. Within that, three suppliers sit above 8% share individually, and the top supplier sits at 12%. The category-level view shows Packaging at 62% concentration with a single strategic supplier; IT Services at 34% with two suppliers dual-sourced; Logistics at 89% with one supplier effectively single-sourced.

The tail-spend view shows 340 suppliers representing less than 0.1% of spend each, in aggregate accounting for 7% of total spend. Within that tail, 140 suppliers are used for categories where a preferred-supplier alternative exists, which is the consolidation hypothesis.

The maverick spend analysis flags 4.2% of spend as potentially off-contract: purchases made through non-preferred suppliers in categories where a preferred supplier contract exists. The flagged lines include supplier names, amounts, and the applicable preferred-supplier contract for each case.

The savings hypothesis: tail-spend consolidation represents a 2-3% recovery against the consolidated spend; maverick spend recapture represents a further 1-2%; selective renegotiation of the high-concentration categories could produce 3-5% in the packaging and logistics spend. Combined hypothesis: €4-7 million of addressable savings against the €180 million base.

The procurement analyst's next job is not to produce the analysis. That is done. The analyst's next job is to validate the findings, test the savings hypothesis against market conditions and commercial relationships, and prioritise the opportunities into a savings roadmap. That is the strategic work, and it is the work Copilot did not displace.

What Copilot in Excel Does Not Do

Three limits worth naming.

Copilot does not replace category expertise. A savings hypothesis is a starting point, not an answer. Whether 3% is achievable on tail-spend consolidation depends on the specific supplier relationships, the category dynamics, and the contracting cycle, which the procurement analyst understands and Copilot does not. The analysis accelerates the mechanical work; it does not remove the commercial judgement.

Copilot does not replace supplier master data management. Dirty data produces dirty analysis, Copilot or no Copilot. Procurement organisations that want to run a monthly spend refresh reliably need a supplier master data foundation. Copilot makes the foundation visible faster; it does not build the foundation.

Copilot does not replace validation. Every material number in the analysis should be spot-checked against the source data. A top-ten supplier that Copilot has miscategorised, or a concentration percentage that is computed against the wrong denominator, can invert a conclusion. The validation is fast, ten minutes for a typical analysis, and it is essential. Procurement analysts who skip validation produce reports that get challenged in the first review and erode the credibility of the whole AI-assisted approach.

Procurement teams building this into a real monthly rhythm, and there are more of these now than a year ago, typically combine Copilot-assisted spend analysis with a structured training programme that covers the prompt design, the validation disciplines, and the category-specific interpretation skills. The AI Fundamentals for Procurement Teams program is built around exactly this kind of integration of AI tooling with procurement expertise.

Where Copilot in Excel Fits in the Broader Procurement AI Stack

Procurement teams running multiple AI tools tend to give Copilot in Excel the spend work. Claude or ChatGPT might also be able to run spend analysis, but Copilot's Excel-native integration means the work stays inside the spreadsheet, no copying outputs into the workbook, no reconciliation between Copilot's interpretation and the actual file, no version-control question about which copy of the spend file is authoritative.

For long contract review, supplier risk analysis across external market data, and complex category strategy work, Claude or a dedicated tool often does better. For everything that starts and ends inside an Excel workbook, spend analysis, scorecards, forecasting, KPI dashboards, Copilot is usually the right default. The procurement team's job is to get the workflow-to-tool mapping right, not to force every workflow onto a single tool.

Frequently asked questions

How clean does the spend data need to be?

Consistent supplier names, single currency, consistent category taxonomy, defined time period. A Copilot prompt can produce suggested cleanup as a first step, but the best outputs come from cleanup performed before the analysis runs.

How do I validate Copilot's numbers against the source data?

Spot-check the top-line totals (total spend, top-ten share, tail-spend share) against pivot tables on the source data. If the totals reconcile, the detailed output is usually reliable. If they do not, the source data likely has a structural issue, multiple currencies, duplicate rows, or inconsistent categorisation, that needs to be resolved before the analysis continues.

Is this workflow appropriate for sensitive supplier commercial data?

Yes on an enterprise Microsoft 365 Copilot plan with the standard data-handling terms, because the data stays inside the organisation's Microsoft 365 tenant. The AI policy should still specify which categories of spend data are appropriate for Copilot-assisted analysis.