“This will replace the spreadsheet…”

I first heard that pitch years ago, when the company I worked for still ran on Lotus 1-2-3. It became a familiar refrain from the financial software industry over the decades that followed. Lotus did get replaced, by Excel, but every attempt since to take a meaningful bite out of Excel’s territory has n’t worked.

Even Microsoft hasn’t managed it. Its newer tools have widened Excel’s moat rather than drained it. Power BI is a poor starting point for a financial forecast, and the tools that would make it work, Power Query and Power Pivot, sit inside Excel itself. The humble spreadsheet has grown beyond its 2D grid into something much closer to a full-blown software application.

That shift has put the burden on finance professionals to learn the tools. The power at their fingertips is immense, yet I’ve worked with plenty of finance people who have got through most of their careers on VLOOKUP. The result is spreadsheets that swallow vast amounts of time and still leave the user ferreting through worksheets for anything useful.

AI is changing this. Back in the mid-90s (a clue to my age), any mention of AI would more likely have come from a farmer worried about fertility in his herd. Today, finance professionals can use LLMs to build small, discrete applications that take the automation and data wrangling out of Excel. These tools sit alongside the input file and turn it into a clean, transformed dataset. That output can feed Excel for the summary reports or go upstream to Power BI or a data portal on a website.

This approach brings real advantages, and they start with the spreadsheet’s biggest flaw. Its greatest strength is also its greatest weakness: flexibility. That flexibility breeds poor practices: data input sections scattered throughout the file, unprotected formulae, and a game of hunt the thimble every time you try to trace a number back to its source. I could go on, and no doubt you could add to the list.

For the past two years I’ve been building discrete applications that tackle the most common pain points in a finance team: reporting, forecasting and reconciliations. I’m not a software developer. I’m an accountant who taught himself to build these tools, and AI is what made that practical. e.g. stock reconciliation model that used to take two days now runs in minutes. The key is to leave the spreadsheet doing what it does best: dynamic reporting that’s easy to use and quick to build.

This isn’t another pitch to replace the spreadsheet. It’s about taking the heavy lifting out of it, so it can get back to doing the job it’s good at.

Over the coming articles I’ll go into the detail: how I’ve built these tools using AI, what worked, the pitfalls, and the tips I wish I’d had at the start.