Related sites:
Newsletter: Perspectives on Power Platform
Company: Niiranen Advisory Oy
Happy 40th birthday, Microsoft Excel!🎉
I wish I could start this post by “the first time I used Excel…” But that’s impossible since it feels like Excel has always been there. Not on my very first 386sx PC in the early 90s, no. Back then I was mostly doing games, music, and the occasional pre-internet discussions in different forums (my previous post is all about that). When I started my business studies and when I got my first job roles, though, Excel was just how things got done. A lot of things.
When I got the chance to visit Redmond for the first time as part of the Microsoft MVP Summit, of course I bought a souvenir from the Microsoft company store that was about Excel. Today, I’m still extremely happy to wear this beautiful shirt:
The significance of Microsoft Excel and spreadsheet software in general can hardly be overstated. It is truly the nr. 1 business app. Many people could easily do without the other MS Office products, replacing them with Google’s apps or whatever modern tools have emerged in the past couple of decades. But Excel? So many folks would simply hate having to use anything else.
Why is Excel so unbeatable? Why do so many companies try to fight it nevertheless? Let’s reflect on that.
Trying to fight Excel
I started working with customer data analytics and direct marketing at the turn of the millennium. Back in 2000, there was no SaaS. Instead, companies invested in server software and hardware to do things that today’s Excel workbooks with 1 million row limit could easily handle. This scarcity of apps also meant that everything which did not have a dedicated server + client combo in place was done with Excel.
The Holy Grail of “Customer 360” required a central database for customer information, which meant years/decades of trying to migrate the Excel-based business processes into CRM systems. At the start of such a journey, everyone was quick to agree that “yeah, this thing shouldn’t really be in Excel workbooks”. Once a system was in place, however, the question turned to “can I export this info into Excel?”
“In the analytics industry there is a standing joke that “Export to Excel” is the most used feature of any analytics software.”
Hjalmar Gislason: Export to Excel — business software’s most common feature?
Why do people want to go back to the world of spreadsheet after they have deployed a purpose built business application for the data in question? Because FREEDOM!!!
Sure, the apps that are designed for specific business processes can give readymade views into the data that you can open with a click. They usually also offer automation that could perform actions like data updates, notifications, summaries, integration, and all sorts of tasks that would be much more laborious when working with spreadsheets.
It all sounds incredibly useful, especially before you have any such apps in place. The idea of being able to rely on a tool from developers that have already figured out what you’d want the computer to do is just beautiful. Everybody wants to spend less time thinking about “how do I do this” and would rather have “an app for that”. This has given birth to countless SaaS solutions that are simply unbundling Excel spreadsheets into dedicated apps:
And yet none of the apps can escape the “I wish this worked like Excel” request from the users. Inevitably there will come a time when either the features of the app feel like a constraint that gets in the way – or the users just can’t figure out the intended process of how to do things in this specific, unique application that they’ve not used for this purpose before.
Why Excel rocks
For a business applications guy like me, it’s natural to see a process managed with Excel and immediately jump into thinking what data models, interfaces, automations and integrations could be put in place to make everything more efficient. Starting from CRM, the assignments I have been given usually always include the “M” – managing things. Not just for a single PC user but for a group of users, sometimes spanning to thousands of individuals.
Business apps with a clear separation of the server-side database and logic from the client-side UI are indeed a better way to managing things involving many users. Yet we often forget that it’s not enough to cover 80% of the tasks with built-in features of our slick, modern web apps. The employees who are in charge of end-to-end processes must complete 100% of what their role requires. If our minimum viable product doesn’t have complete coverage, the users need to figure out a way to get the remaining work done.
That’s where Excel is the hero. It adapts to the needs of the user unlike anything else available to typical information workers. Whereas Word documents are often just digital versions of paper docs, Excel introduces dimensions not found from the physical office tools. Its UI invites the users to model their problem domain in a 2D digital canvas – much like a box of Legos will invite kids (and grown ups) to build abstract plastic versions of real-world things.
Mr. Alan Cooper, the father of Visual Basic, calls this the fudgability of software. “[Excel is] a terrible program, but it’s powerful, flexible, and it allows its users to work fudgably, adaptably, in real time, while seeing most of what is happening right there on the screen.”
Sometimes we need rigidity in business processes. We want to have repeatable outcomes that aren’t different based on who is operating the tools, or where & when. Business apps are excellent for facilitating a common way of working. At the same time, we must remember that a significant portion of information work consists of ad-hoc requirements and fuzzy processes adapting to incomplete data.
That’s when we rely on the humans to understand what should be done. They, in turn, will rely on available tools that allow them to tackle data processing needs without unnecessary limitations. Meaning spreadsheets in most cases. Yeah, I wish more people were able to turn those repeatedly used Excel sheets into low-code solutions with Power Platform tools, yet that’s a bit too much to ask from individual workers most of the time. Most often, Excel is just a better fit.
LLMs only wish they could be like Excel
AI is supposedly eating the software world. It certainly is eating all of the available capital (and more) in trying to turn computers that used to be reliable into something that’s… not. The advantages of having the computer understand natural language and being able to respond to any question are of course huge. If only there wasn’t that minor inconvenience that we can’t know if the response is correct or not.
Microsoft, a.k.a. The Copilot Company, has of course been fearlessly approaching the idea of combining creative large language models with software that is normally used for precise calculations. The result is news headlines like this: Microsoft launches Copilot AI function in Excel, but warns not to use it in ‘any task requiring accuracy or reproducibility’. Okay, cool, I only ever used Excel for fun and games anyway, so no worries about those accuracy gaps…
The area where creativity and conceptual problem solving could be very useful is in designing how to get from an Excel into an app. I’m optimistic about the LLM-based Maker features accelerating and expanding the possible Power Platform solutions that can be built for varying types of business problems. If we don’t expect AI to work the same way every time, but rather use it as a force multiplier for fudgability, the risks of hallucinations breaking the business processes should be much lower.
Let’s look at one case where I’d like to see AI replace Excel in my personal life. For 4 years now, I’ve been tracking the expenses related to my car using an Excel workbook. Fuel amount, cost, mileage are an example of what I’ve entered there after every visit to the gas station:
Many, many times, I’ve thought to myself “I should turn this into an app”. After all, what kind of a Power Platform evangelist would just keep working on an Excel file for years without modernizing the process? Yet whenever I started to think about the details of what should be there to replicate everything my workbook provides (there are other sheets there, too), I realized it’s a heck of a lot of work to achieve parity with what Excel gives me. No rational reason ever existed for me to put in all the work to achieve something that I’d be happier using. It could have been a community demo, sure, but I always had other stuff to work on.
Now, when we are living in the age of agentic AI, surely it’s about time to replace the Excel? Well, if I could just give that file to an AI tool and make the machine modernize it for me, why not. Last week, when Lovable announced their file uploads feature and claimed “you can now drop files directly into Lovable and turn them into apps and websites”, I decided to try it out. Sure enough, the vibe-coding tool was able to generate a web app from the .xlsx file:
The only problem? The data is not correct. Like, I only wish my 2019 Ford Focus 1.5 EcoBoost was able to run 100 km on 0.8L… In reality, the React app generated by Lovable was not able to handle the “big data” of around 200+ lines of data in the Excel workbook and instead chose to read only parts of it. The AI chatbot of course will claim this app has “real calculations from your Excel”, but we all should know at this point how LLMs are just manipulation machines that you shouldn’t spend time arguing with.
I never have to worry if Excel is trying to manipulate me. It may not always understand what I’m trying to do, yet the machine never pretends that it did the work it promised to do while giving out a bogus number instead. The math is not based on vibes. Today, we have LLM-based AI features inside our spreadsheets that are sometimes able to call deterministic tools like Python to do real calculations. But we have no certainty that AI would always resort to such tools when needed, instead of just making shit up.
That doesn’t stop companies from trying to make AI in Excel work. OpenAI has funded startups like Endex that develop Excel add-ins for injecting LLM magic into spreadsheets. Microsoft is also aiming beyond Excel Copilot with their recent preview launch of Agent Mode in Excel. Could these tools already replace a financial analyst creating Excel workbooks with formulas and business metrics? I decided to give Agent Mode a try on the day it came out:
You can check out the experience and end results from the video. In short: Excel Agent Mode ain’t necessarily a tool you should yet rely on…
If you can’t beat them, join them
Instead of seeing Excel as something that should be replaced, perhaps a better strategy is trying to accept the fact that Excel files will always exist. Because people want to experience the control that Excel gives them, instead of limiting themselves to the GUI prison of your business app. Similarly, they’d prefer if the numbers that get shown to them would be based on verifiable math rather than on-demand hallucinations of “vibe working” tools.
I’m somewhat biased here because this is the route we have taken when building FinModeler. A SaaS application that helps business founders or financial advisors to create detailed financial models with its simple web-based wizard UI. And then, producing an Excel workbook for you, complete with dynamic formulas that allow you to adjust the model with the tool you know and love.
It’s not your typical product built on Power Platform. Not simply because of the full-fidelity Excel workbook generator feature. But rather the whole SaaS delivery model of offering a canvas app that requires zero installation differs from the expected way of partners shipping products on Microsoft’s low-code platform. In many ways, this is not how business apps are supposed to work – but we’ve done it anyway!
Today, anyone can check out how this combination of Power Platform + Excel works in practice by signing up for a free 14d trial of the FinModeler app. There’s a lot more that we can do in future versions of the product, besides just creating extensive Excel workbooks. The fact that the financial model data is stored in a structured Dataverse database is a much better position to build new features than if we’d need to rely on Excel files alone. Still, it’s essential that the users have the possibility to interact with the model that offers them the ultimate level of control and confidence.
Spreadsheets are forever
The moral of the story is: it’s not either/or. We are all better off when there are different types of tools available for working with business data and processes. Both structured apps and fudgable spreadsheets serve a clear purpose. Similarly, we have room for both deterministic, reliable software as well as creative and unpredictable AI. Trying to force people into choosing just one tool never works.
Recent estimates say Excel is used by over a billion human users. In the near future, there will be countless AI agents built that will also try their best to learn how to work with Excel spreadsheets. What I’ve learned is that while you can (and often should) go beyond spreadsheets to evolve and improve your processes, it’s foolish to think that you could replace Excel entirely. Turn it around instead and think of all the data manipulation and calculation features you don’t have to build into your own software, thanks to Excel being there to handle it with ease.
After 25 years of working professionally in the field of customer data and business applications, I’m proud to once again add a sticker onto my laptop that celebrates the magical powers of Excel:
