What is Sheets AI and what does it do in Google Sheets?

Sheets AI refers to the Gemini-powered AI features in Google Sheets, available on Google Workspace Business Standard ($14/user/month as of 2026). It includes formula suggestions, conversational data analysis, and the "Help me organize" feature for structuring flat data into tables. The context window is capped at approximately 25,000 cells per query, and all analysis is scoped to the active sheet — it has no cross-tab awareness.

Can Sheets AI write formulas that reference other tabs?

Not reliably. Gemini's formula suggestions work well for single-tab lookups and common aggregations, but they don't reliably produce cross-tab references like `=SUMIFS('P&L'!C:C,'P&L'!B:B,">="&Assumptions!$B$3)`. You can get there by prompting very explicitly with sheet names and ranges, but it won't infer cross-tab structure on its own. For multi-tab formula writing, you'll need an assistant with full workbook context.

Does the 25,000-cell limit affect real financial models?

Yes, frequently. A 1,000-row SKU-level P&L with 15 columns is 15,000 cells — under the limit on its own. But combine that with a prior-year comparison range and you're over. Anything with dense transaction-level data (actuals from a GL export, SKU-by-region contribution margin tables) will hit the ceiling regularly. Google's Gemini in Sheets documentation confirms this limit as of April 2026.

Is Sheets AI worth it for FP&A teams if they're already paying for Google Workspace?

If your team is on Business Standard or higher, you already have access — there's no incremental cost to evaluate. The formula suggestion feature alone saves time on standard aggregations and is worth the 5 minutes to try. The honest ceiling is that it won't help with cross-tab model logic, consistency checking across a linked 3-statement model, or anything where the relevant data spans multiple sheets. McKinsey's 2024 AI survey found fewer than 20% of finance teams describe their AI tools as integrated into core modeling processes — Sheets AI reflects exactly why.

What's the difference between Sheets AI and a dedicated spreadsheet AI assistant?

The core difference is context scope. Sheets AI sees one tab at a time, up to 25,000 cells. A dedicated assistant like ModelMonkey maintains context across your full workbook — all tabs, their structure, and how they reference each other. For board pack prep or DCF work where the WACC flows from Assumptions into the FCFF tab and then into Returns Analysis, cross-tab context is the whole job. Single-tab AI won't trace that chain.

Google Sheets AI: What FP&A Analysts Actually Get

This is an honest assessment of what the native AI features in Google Sheets do, where they break down on serious modeling work, and what the tradeoff looks like for a team running board packs, DCFs, and sensitivity analyses on linked multi-tab models.

What Sheets AI Can Actually Do Right Now

Google's Gemini integration in Sheets (part of Google Workspace Business Standard at $14/user/month as of Q2 2026) ships with 3 meaningful capabilities:

Formula suggestions. Type a description of what you want and Gemini proposes a formula. Works well for common cases — SUMIFS, VLOOKUP, basic array formulas. Falls apart on anything involving structured references across named ranges or multi-tab lookups. It won't reliably write =SUMIFS('P&L'!C:C,'P&L'!B:B,">="&Assumptions!$B$3) without a lot of hand-holding.

"Help me organize." Converts flat data into structured tables, suggests header names, infers data types. Useful for ingesting raw exports from a GL or Stripe. Not useful for anything your model already has structure on.

Conversational data analysis. Ask a natural language question about data in the current sheet. Gemini summarizes, identifies trends, and writes SQL-style queries against your data. The constraint is significant: according to Google's Gemini in Sheets documentation, the AI context window is capped at approximately 25,000 cells per query. On a 1,000-row SKU-level P&L with 15 columns, you're already hitting that ceiling.

Response latency runs 3–5 seconds for moderately complex queries, which is fine for one-off questions and painful when you're iterating.

Sheets AI for Multi-Tab Financial Models: The Honest Evaluation

Here's where it gets blunt: Gemini in Sheets has no awareness of cross-tab relationships.

Your 3-statement model has 8 tabs. The FCFF tab pulls from the P&L. The Returns Analysis tab pulls from FCFF, the Assumptions tab, and the Balance Sheet. Gemini sees one tab at a time. Ask it to "check if my EBITDA flows correctly into the DCF" and it can't — it doesn't have the cross-tab context to trace that linkage.

The practical failure modes:

Formula suggestions don't cross tabs. It won't suggest ='FCFF'!$D$14 when you're building the Terminal Value section of your DCF.
Data analysis is sheet-scoped. Ask "what's driving the variance in my Q3 gross margin?" and Gemini can only answer if your entire P&L history and assumptions live on the same sheet. They don't.
No model-wide consistency checking. It can't tell you that your Assumptions!$B$3 start date is inconsistent with the period labels in your P&L tab.

According to McKinsey's 2024 Global Survey on AI adoption, 60% of finance functions report using AI tools in at least one workflow — but fewer than 20% describe those tools as integrated into their core financial modeling processes. The gap between "AI in finance" and "AI that understands financial models" is exactly where Sheets AI currently sits.

This isn't a criticism of the product. It's a single-tab AI assistant that does single-tab things well. The problem is the marketing doesn't say that.

Where the Gap Shows Up in Practice

Three scenarios where Sheets AI's ceiling hits hard:

Quarterly board pack. You're pulling actuals from the P&L, comparing to budget on the Assumptions tab, and summarizing on the Executive tab. Gemini can analyze each sheet individually. It cannot synthesize across them without you copy-pasting data into a single range first — which defeats the point.

Runway sensitivity on new hire pace. Your headcount assumptions live in one tab, the cash burn calculation in another, the runway output in a third. Asking Gemini to "run a sensitivity on hire timing" returns a blank stare unless all three tabs are collapsed into one.

Contribution margin by SKU with YoY comparison. If your SKU-level data and prior year data are on separate tabs — standard practice to keep the model readable — the 25,000-cell limit and the single-tab constraint collide simultaneously.

The formula suggestion feature is genuinely good at the 85% case: common formulas, single-tab lookups, basic aggregations. It saves time on work that wasn't the bottleneck anyway.

What Actually Helps Multi-Tab Models

The gap Sheets AI can't close is cross-tab context: understanding how data flows from Assumptions through the P&L and into a Returns tab, catching inconsistencies, and suggesting formulas that reference the right sheet.

ModelMonkey is built for exactly this. It sits inside Google Sheets as an add-on and maintains context across your full workbook — not just the active tab. Ask it to write a SUMIFS that filters your P&L by the date range in Assumptions and it knows both tabs exist, what columns they have, and how they're linked. It can check whether your FCFF calculation correctly excludes working capital items that appear in the Balance Sheet. That's the difference between a single-tab assistant and something that understands model architecture.

For teams running linked 8-tab models where a number mis-tying costs credibility, that distinction matters.