Your Spreadsheet Isn't Broken. Your Process Is.
The close doesn't have to take this long. Neither does the forecast, the board deck, or the variance hunt.
created by ChatGPT Image 1.5
Excel just got a real AI co-worker
If you’ve ever lost an hour to a broken VLOOKUP, a messy export, or a workbook you didn’t build, you know the feeling.
Excel isn’t hard because the math is hard. It’s hard because the work is brittle. One weird date format, one hidden filter, one “temporary” hard-coded number, and the whole thing turns into a puzzle.
That’s what’s changing right now.
Tools like ChatGPT 5.2 (especially in deeper reasoning mode), Claude Opus 4.5, Codex, Claude Code, and Claude for Excel are getting genuinely useful at Excel-shaped work. Not “here’s a formula suggestion.” More like: “I can read the workbook, understand what you’re trying to do, propose a clean approach, and help you implement it without breaking everything.”
The shift executives should pay attention to
For years, AI in spreadsheets meant autocomplete. Helpful, but small.
Now it’s becoming an actual layer of work.
These tools can take a messy workbook, figure out what it is, and help you move from question to output. Formulas that handle edge cases. Pivot tables with calculated fields. Charts that tell the story. A short written summary that matches the numbers.
That last part matters more than people think.
Because Excel isn’t “just a tool.” It’s where finance teams run the business, where ops teams report truth, and where decisions get made when the systems don’t line up.
If AI can reduce the time it takes to get from raw data to a trusted answer, that changes cycle times across the whole company.
Why this is critical for finance and leadership
Most executive teams feel the same pain, even if they describe it differently.
The close takes too long. Forecasts get rebuilt from scratch. Variance analysis turns into a scavenger hunt. Board decks are a week of manual work. And only two people really understand the model that everyone depends on.
That’s not a “spreadsheet problem.” That’s a speed and control problem.
When your finance org spends half its time cleaning data, fixing links, and explaining the same logic over and over, you don’t get better decisions. You get slower decisions. And the business runs on stale numbers.
The promise here isn’t that AI makes Excel prettier.
It’s that AI can compress the time between “What’s happening?” and “Here’s the answer I trust.” And it can do it while reducing key-person risk, because the logic becomes easier to explain, document, and repeat.
Five high-value Excel workflows where this pays off fast
Here are the places I see the biggest upside, especially in FP&A and finance operations.
Variance analysis that doesn’t take days
Imagine dropping in your actuals export and your budget, then asking the tool to build the variance table, create a simple bridge chart, and draft a tight explanation: what moved, why it moved, and what’s likely to persist. Not a novel. Six to eight sentences you’d actually send to leadership.Forecast refresh without rebuilding the world
Many teams treat forecasting like a monthly rebuild because the model is fragile. A good AI assistant can help you update assumptions, check what broke, and run scenario tables without you babysitting every formula. “Show me base, downside, and upside with headcount and churn sensitivity” becomes a normal request, not a weekend project.Cash and working capital visibility that stays current
Cash forecasts often die because the inputs are messy. Collections timing, vendor payments, payroll, seasonality. AI can help clean the feeds, identify missing items, and keep the structure consistent so finance can spend time on the drivers, not the mechanics.Pricing, margin, and deal analysis that’s usable
The deal desk version of Excel is usually a pile of tabs and assumptions. AI can help standardize it. Create a margin waterfall. Show break-even points. Flag deals that look good on revenue but bad on cash or delivery capacity. That kind of analysis is how you protect profit, not just growth.Board and QBR packs that stop being hero work
A lot of leadership reporting is the same every month: KPIs, trends, a few charts, and commentary. AI can generate the first draft of that work, then you review and adjust. That shift is huge. It turns the pack from “two analysts and a prayer” into something repeatable.
None of this replaces judgment. It removes the grind that keeps judgment from showing up on time.
And it’s not just finance
Sales ops can turn pipeline exports into clean stage conversion views, segment performance, and “what changed since last week” summaries.
Marketing teams can do cohort views, CAC trends, and channel mix reporting without wrestling the same pivot tables every month.
HR can run headcount planning, attrition views, comp bands, and manager rollups without a tangle of ad hoc sheets.
Procurement can compare vendor pricing, normalize line items, and track renewals with less manual cleanup.
The pattern is the same everywhere. The work isn’t hard, it’s tedious and fragile. AI is getting good at tedious and fragile.
Where each tool fits
ChatGPT 5.2 is great when you want a fast analyst and a clear explainer. Upload the workbook, ask for the story, ask it to sanity-check the outputs, and get formulas, charts, and a clean write-up you can paste into an email.
Claude Opus 4.5 shines when the spreadsheet is a model, not just a table. Forecasts, sensitivities, scenario grids, and multi-step edits that need to stay consistent across tabs are where it tends to feel strongest.
Claude for Excel is about staying in the sheet. It’s for the person who lives in Excel all day and wants help right there: explain this tab, rewrite this formula safely, build a pivot, create a chart, and document what changed.
Codex and Claude Code are for repeatable workflows. If your “Excel work” starts as files in a folder and ends as an updated workbook every Monday, these tools help you turn a manual routine into a script. Not fancy. Just dependable.
If you’re deciding what to roll out, here’s the simple rule: in-sheet assistants help individuals. Code-first assistants help teams build repeatable pipelines.
Most companies will end up using both.
The new skill is working in loops, not one-shot prompts
The biggest mistake people make is asking for a final answer immediately.
The better approach is a short loop:
Start with: “Tell me what you think this workbook is doing. What are the key tables. What assumptions do you see. What looks fragile.”
Then: “Propose the cleanest way to answer this question, and list the steps.”
Then: “Make the changes and give me a change log. Tell me which cells you edited, which formulas you added, and what checks you ran.”
This sounds small. But it’s the difference between “AI as a toy” and “AI as a reliable junior analyst.”
The guardrails finance leaders should insist on
AI can still make a confident mistake. So the right posture is “draft, then verify,” not “paste and pray.”
For finance teams, a few controls make this safe and scalable:
Treat every AI-modified workbook like a changed model. Keep versions.
Require check totals. If the totals don’t tie, nothing ships.
Lock critical tabs and inputs where possible. Reduce accidental edits.
Ask for assumptions and exclusions in plain English. Make them visible.
Don’t mix sensitive client or employee data into casual tools. Use approved environments and permissions.
If you do this well, you get speed and control. That’s the prize.
What to do Monday morning
Pick one recurring Excel task that burns real hours every week.
Define “correct” with two or three check totals or known outputs.
Run it with one tool and save the exact prompt you used.
Add a review step you will always repeat before sharing results.
Turn it into a template and teach it to one other person on your team.
Excel isn’t going away. But the time you spend fighting it can.
And if you run finance, this is bigger than personal productivity. It’s an opportunity to shorten decision cycles, improve trust in the numbers, and free your best people to do the work you hired them for.
Why I write these articles:
I write these pieces because senior leaders don’t need another AI tool ranking. They need someone who can look at how work actually moves through their organization and say: here’s where AI belongs, here’s where your team and current tools should still lead, and here’s how to keep all of it safe and compliant.
In this article, we looked at where Excel work gets stuck - not because the math is hard, but because the process is brittle - and how AI tools like Claude Opus, ChatGPT, and Codex are crossing a threshold where they genuinely help. The market is noisy, but the path forward is usually simpler than the hype suggests.
If you want help sorting this out:
Reply to this or email me at steve@intelligencebyintent.com. Tell me where your finance or ops team is losing hours to manual spreadsheet work and what “correct” looks like for those outputs. I’ll tell you what I’d test first, which part of the Claude/GPT/Codex stack fits, and whether it makes sense for us to go further than that first conversation.
Not ready to talk yet?
Subscribe to my daily newsletter at smithstephen.com. I publish short, practical takes on AI for business leaders who need signal, not noise.