Why Excel Will Soon Die

Excel is one of the greatest productivity tools ever created. But I believe we are now entering a new stage, what I call removing the input layer.

Lower costs, fewer errors and thousands of hours saved. If software can process financial information using deterministic, predictable rules, with near-perfect accuracy and in a fraction of the time, what is stopping it? See the following facts:

• Python libraries such as pandas can read entire Excel workbooks, filter thousands of rows, apply accounting rules, calculate new values and write the results back into another spreadsheet
• At the simplest level, email attachments can be downloaded or dragged into a Python program for processing. Going a step further, workflow platforms such as n8n and Microsoft Power Automate can monitor incoming emails, retrieve attachments automatically, move information between applications and update spreadsheets without manual intervention
• Robotic-process-automation tools can even operate older systems by reproducing the same clicks and data-entry steps currently performed by an employee, using simple tools like AppleScript or Microsoft Power Automate

Previously, the main obstacle was the technical barrier. Building these automations required skilled engineers, a large budget and a lot of time. But that is changing.

Two things are driving this shift: (1) the democratization of open-source automation frameworks and, more recently, (2) the democratization of coding itself through AI. The uncomfortable reality is that many finance and administrative roles still involve hours of copying information into Excel, moving it between systems and repeatedly applying the same rules. Excel was an enormous improvement over handwritten ledgers and physical logbooks; however, I believe the next phase will remove much of that manual input, along with many of the roles built around it.

Removing the input layer

Every financial process begins with an original source of information: an invoice, an email, a receipt, a schedule, a bank statement or another spreadsheet.

We often assume that the person handling this information is creating value by typing the figures into Excel or applying a set of accounting rules before entering them. Excel then performs the formulas, updates the totals and generates the final output.

But in many cases, the human is not making a genuinely new decision. They are simply acting as the input layer between the original information and the spreadsheet.

If the rules can be clearly defined, a program can read the original source, interpret the relevant information, apply the same accounting logic and populate the Excel file automatically.

Any deterministic process can be automated

Excel functions are useful because they are deterministic: when given the same input, they produce the same output.

The same principle applies to many accounting and administrative workflows. If a process follows a clear and repeatable set of rules, much of it can usually be automated. This input rule applies not only to Excel, but also to QuickBooks, Xero, enterprise resource planning systems and almost any other piece of business software. Even tiny actions, a button clicked, a checkbox ticked or a value copied, add up when repeated even with a single employee across thousands of transactions.

Here are two examples from my experience automating financial workflows at a multinational BPO company.

Case 1: Excel sheet to Excel sheet

The first case involved transforming an accounts-payable ageing report into a vendor-payment request. An administrator previously had to review each vendor section, identify which bills qualified for payment, exclude irrelevant entries and manually transfer the required information into a new spreadsheet.

Although the report appeared complex, most of the work followed a fixed sequence:

Because these decisions followed clear rules, the entire transformation could be automated.

Case 2: Capacity planning, email to Excel

The second case went further because the original information came directly from an email.

Every week, employee schedules arrived as an HTML table. An administrator had to read each entry, identify the employee's day and shift, locate the correct row and time slot in a large Excel calendar, and enter the information manually.

Because the process followed predictable rules, the email could be interpreted and the workbook updated automatically. Instead of inspecting every row, employees only needed to review the unusual cases the system could not confidently determine.

This is an important principle: automation does not require every input to be perfectly clean. A well-designed system automates the “predictable majority” and escalates only the exceptions. Humans are then able to focus on the small number of cases that genuinely require judgement.

How about handwritten or printed invoices?

You may think one of the remaining situations in which someone may still need to manually type information into Excel is when the original information comes from a physical document: a receipt, bank statement, invoice, official letter or handwritten form.

However, even this barrier is disappearing. Recent advances in optical character recognition allow software to extract information from increasingly difficult documents, including poorly formatted scans and messy handwriting, with one example being Baidu AI's Unlimited OCR project. The broader point is not that every document can already be processed perfectly. But the point is that the number of processes requiring manual encoding continues to shrink.

In fact, in a few years, within the AI era, the same transition will happen not only with Excel, but with almost every form of human-software interaction.

Rapid developments in agentic AI will make it much easier to understand and redesign company workflows. People will no longer be valued simply because they can enter information quickly into Excel, QuickBooks, Xero or another specialized platform. Once a reliable system is built, much of that manual work no longer needs to be repeated.

What will matter is the ability to understand information and help businesses make sound decisions. At its core, value will come from solving problems, not simply “operating software.”

1. Businesses need to handle this transition responsibly. I believe that companies are truly made up of people, and that most companies do care about their employees and would rather help them grow than leave them behind. The best-case scenario for both the company and its employees is to act early. This means redesigning workflows and retraining employees into higher-value tasks, such as financial analysis and stronger audit control.
2. Automation should not simply remove people. But the danger comes when companies wait too long. Once competitors automate, lower their costs and become more efficient, slower companies may be forced to automate urgently just to survive. Removing roles may be the unfortunate option, the sad reality.

Moving forward

I was 17 when I worked on these systems. By my second week, I was already proposing and implementing changes to existing finance processes. This isn't a testament to my ability, I was a 17-year-old intern, after all, but it's proof that learning and iterating fast is more than possible.

During this internship at a large multinational BPO, I expected pushback. But instead, they had this culture of continuous improvement and were more than willing to listen. That openness is the difference: companies that stay agile and act on automation will outpace those slowed by bureaucracy. Oftentimes, the technology is not always the hardest part. The harder part is being willing to look at an existing process and ask: why are we still doing it this way?

AI is now lowering the barrier to building automated workflows. Accountants, administrators and managers can increasingly describe a process, build an initial system and refine it without even being actual engineers. There are already accountants using tools such as Gemini and Lovable to build expense systems, financial dashboards, close-management tools and other internal applications.

Building these systems is becoming far more accessible. Very soon, creating automated workflows will not be limited to technical teams. It will become a normal part of how businesses improve.

The most valuable people will not be those who simply follow a process, but the ones who understand it, question it and make it better. And I believe that person can be any one of us.