The problem with your budget starts at “Export to Excel”
In October 2020, Public Health England lost track of 15,841 COVID-19 cases. There was no hack and no system crash. The data had been moved into an old Excel file that ran out of rows, and everything past the limit disappeared without a warning.
That’s the dramatic version. The everyday version is sitting in your shared drive. A review of 35 years of spreadsheet research found that 94% of spreadsheets used in business decisions contain errors. Most of them never make the news. They just make it into the budget.
And a lot of them start the same way: someone clicks “Export to Excel.” At our recent Budgeting in the Era of AI workshop, Natasha Sabelli, CPA and Velixo’s Finance Evangelist, described what happens next. The moment ERP data lands in a spreadsheet, it becomes “a disconnected snapshot,” and Excel turns into “an island of stale data.”
What breaks first: the trail back to your ERP
In the ERP, every number has a history. You can see who posted it, when, and what it rolls up to. Once it’s pasted into a workbook, that history is gone. It’s just a number in a cell.
That’s fine until someone asks where it came from. Three months later, the CFO wants to know why Q1 marketing spend in the board deck doesn’t match the ERP, and finance starts digging. Which export was it? Which version of the workbook? Did anyone pull the numbers again after the Q1 adjustments posted? It turns into a scavenger hunt through shared files and old Teams and Slack messages, and it’s the reason version control topped Natasha’s live poll for worst part of budget season.
The bigger cost is trust. When the board deck, the department budgets and the ERP each tell a slightly different story, nobody can say with confidence which one is right. Finance ends up defending all three, and every meeting starts with reconciling the numbers instead of discussing them.
What breaks next: your AI answers
Now add AI to the picture. You open that same workbook and ask, “Are we on track to hit our Q3 revenue target?” A few seconds later, you get a clean answer: revenue is running 4% ahead of budget, driven by two large orders in August, and there’s enough cushion to approve the new equipment purchase a quarter early. It’s clear, confident and nicely formatted. It might even end up in the board deck.
Here’s what it doesn’t know. After you hit export, a customer returned part of one of those August orders, and the credit memo was posted on Thursday and a second invoice was reversed because it had been billed twice. Add those back and revenue isn’t 4% ahead. It’s slightly behind, and the equipment you just approved earlier is now a cash question you’ll be answering at the next board meeting.
To be fair to the AI, it answered the question with the data you gave it. As Natasha put it, AI is “a tool that is only as good as the data you point it at.” A clean answer from a stale file is still a stale answer.
The research backs her up. Gartner predicts that through 2026, organizations will abandon 60% of AI projects that aren’t supported by AI-ready data, and 63% of organizations either don’t have the right data practices for AI or aren’t sure.
The fix starts before the AI does: get your house in order first. In the revenue example, two simple questions would have caught the problem before it reached the board deck: when was this data pulled, and has anything been posted since? Add two more before relying on any AI answer about your numbers: can you trace them back to the ERP, and should this tool be seeing everything in the file? If you’re not sure about any of the four, fix the data before you act on it.
How exposed is your budget?
Here’s the good news: you can check for every one of these problems in an afternoon. Start with the report your leadership team reads every month and count how many exports it takes to build. Search your shared drive for “final” and see how many versions of this year’s budget turn up. Pick one number from last month’s board deck and time how long it takes to trace it back to the ERP. Ask whether anyone rekeys the approved budget into the ERP by hand. And find out whether anyone on the team has already pasted an exported file into an AI tool. (Someone probably has.) Write your answers down. They’re your baseline, and the goal from here is to watch every one of those numbers get smaller.
You can start chipping away at all of this before buying anything. Put an “as of” date on every export, so whoever reads it, human or AI, knows how old it is. Set a data cutoff before budget meetings, so every department walks in with the same numbers. Write the assumption next to the number, so the CFO’s question three months from now takes thirty seconds instead of a scavenger hunt. And give every workbook one owner, so there’s never a budget_FINAL_v14 to explain.
Keep Excel. Connect it to your ERP.
Those habits help, but they’re still workarounds. The longer-term fix keeps Excel right where it is. Nobody needs to give it up, and finance isn’t about to. In the 2025 AFP FP&A Benchmarking Survey, 96% of FP&A professionals said they use spreadsheets for planning. They also named bad data as the biggest barrier to getting value from their technology, ahead of skills or the tools themselves. Excel is where finance does its best work, and the ERP is where the transactions live. The problem is the gap between the two, and every export makes it wider.
Closing that gap is the job of an Excel add-in that connects directly to your ERP. You keep working in the spreadsheet you know, and the numbers stay tied to the source instead of a copy. Once that connection is in place, the budget workflow looks different:
- Actuals flow into the workbook straight from the ERP, with no export.
- You build the budget by department, account and period, with the assumptions written right beside the numbers.
- When something changes, you change it in the same file. There’s no v8.
- Once the budget is approved, you write it back to the ERP, so the system of record matches what finance signed off on.
- You compare actuals to budget all year, on numbers that are current.
Velixo is built for exactly this. It connects Excel to your ERP with a live, two-way connection, so data comes in fresh and approved budgets go back without anyone rekeying them. Yardnique’s VP of Finance, Tom Flanagan, saves two to three hours every time he hits refresh.
It also answers the four questions from earlier. When was the data pulled? At your last refresh, which takes seconds. Has anything posted since? Refresh again and it’s in. Can you trace the numbers back to the ERP? They come straight from it. Should the AI see everything in the file? Velixo Intelligence, now in public preview, works inside the workbook under your own ERP permissions, so it only sees what you’re already allowed to see.
Watch the full Budgeting in the Era of AI workshop
Natasha’s session covers the rest of what has to happen around the model: getting departments speaking the same language, setting the table before templates go out, and documenting the story behind every number. Then Carl Seidman, one of only 35 US-based Microsoft MVPs in Excel, builds a defensible first-pass budget in Part 2.
Watch both sessions on demand and grab the 2027 Budget Kickoff Sheet and the sample workbook to start on your own plan.