KajoData / Excel
Is Excel still worth learning in the AI era? I still start there
I explain where Excel still helps me, how I use it alongside AI, and when I would move from spreadsheet fundamentals to the next analytical tool.

You decide to start learning again, open your browser, and quickly feel as though you have missed everything. Everyone is talking about agents, automation and the next AI capability. You were about to refresh your Excel skills. Is that still a reasonable use of your time?
I think it is. I use AI extensively, and I still value being comfortable in a spreadsheet. Excel helps me organise a problem, show another person what I mean and inspect a simple relationship. I want to learn it for specific work, with a clear point at which I add other tools. I have no ambition to build an empire of worksheet tabs.
I start with the work I want to prepare for
At KajoDataSpace, participants tell me about technical interviews and recruitment exercises. Excel still appears in those accounts. That gives me a concrete reference point. I would not present it as a representative survey of the job market or evidence that every employer wants the same skills.
Before choosing a learning plan, I would look at vacancies I could realistically apply for. I would read the responsibilities, tools, industry and location. I would separate essential requirements from preferences. That is more useful to me than following whichever technology has the most impressive social media post this week.
I would also search beyond one exact job title. Reporting can be part of a role in logistics, finance or customer service. My main question is what problems the person will solve and whether I can already demonstrate some of the necessary competence.
Excel alone cannot guarantee a job. I cannot promise that entering the field will be quick, either. I want my preparation to match an actual starting point. If the work involves spreadsheets, excluding them because another tool attracts more attention makes that preparation harder.
A spreadsheet is a scratchpad for a business conversation
When I say business users, I mean people responsible for prices, deliveries, sales or costs. They may need to decide how much stock to order or explain why a result changed. A report helps them do their own work. They do not necessarily want to spend their day maintaining its technical machinery.
In the source video, I call Excel a scratchpad. That captures an application I value: I open a file, organise a few columns and show how I understand the problem. The other person can immediately check whether I am looking at the right period, product and level of detail.
Imagine a fictional conversation with a warehouse manager. I receive a delivery list and a question about late shipments. First I ask what late means: after the planned hour, after the planned day, or outside an agreed delivery window? Then I build the comparison. A polished table can answer the wrong question if we started with different definitions.
The spreadsheet gives us something concrete to discuss. I can point to one delivery, explain an assumption and show what changes when we correct it. I do not need an elaborate system for this conversation. I need a readable structure and an explanation of the calculation.
There is a limit to the scratchpad approach. If several people copy the file and adjust the rules independently, I lose track of which version is authoritative. A quick sketch should help establish the rules. When the report becomes a regular process, I want to preserve those rules more deliberately.
Fluency still matters when AI is available
Someone in a meeting asks me to change a filter or inspect one customer. I want to make the change and explain it. If every small action requires a fresh conversation with a chatbot, I interrupt the shared investigation.
I do not treat this as a memory competition. I can look up a function I rarely use. There is still a difference between filling one knowledge gap and having no idea what the answer should look like. In the second situation, I struggle to judge whether the AI suggestion fits the task at all.
In the video, I compare that situation to a plumber waiting for phone instructions beside a leaking tap. It is deliberately exaggerated. My point is that I want to perform basic operations consciously, leaving more attention for exceptions and the conversation with the person who needs the result.
I can ask AI to explain where a value comes from or why it suggested a particular way of matching data. Then I inspect the proposal using a small example. My own understanding lets me notice when the tool has answered a slightly different question.
I would not turn this into a universal claim that one person is a fixed number of times faster. Speed depends on the task. My learning goal is more practical: I want to work through an ordinary report and recognise where I need help.
I practise the basics on one small problem
I would begin with a structured table. What does one row represent? Where is its identifier? Which column is an amount, and which is a description? Once that is clear, I would add filtering, sorting, conditional calculations and matching information from another list.
For a fictional exercise, take a list of orders and a separate product list. I want to attach a category to each product using its identifier. Before choosing a formula, I check whether that identifier points to one unambiguous category. Conflicting descriptions for the same key require a decision about the source data.
XLOOKUP is one possible lookup tool. Microsoft’s documentation explains how it returns a value corresponding to a match and notes that it is unavailable in Excel 2016 and 2019. I would check the recipient’s version before choosing a solution. I cover the function in more detail in my XLOOKUP guide.
Next I would add a PivotTable. Microsoft describes it as a way to summarise and compare data and explains the importance of data types and choosing between summing and counting. In the exercise, I want to explain whether the report shows order counts or order values. Those answer different questions.
For me, knowing the basics means I can change a condition and anticipate how the answer should move. I can also investigate a missing match and trace a number back to its source. My Excel learning plan develops that sequence into further exercises.
I would add a request from the report user: show only one category’s orders, while keeping a way to return to the full list. Can I inspect the records included after the change? Can I explain the difference between a missing category and a category with no sales? These small variations make me examine the data instead of replaying the steps from a lesson.
If the solution only works on the unchanged training file, I would try again. I could ask another person to add a few fictional records, or change the task myself the next day. I want to find out whether I remember what the operation means. The appearance of a finished table can give away too much.

I move repeatable preparation into Power Query
Next I look for steps I repeat whenever another file arrives. I remove the same column, correct the date type or select the relevant records. That is where I would introduce Power Query.
Microsoft Learn explains that Power Query records transformations as query steps which can run again during refresh. That gives me a concrete reason to learn it: I can preserve the agreed preparation process instead of recreating it manually from memory.
In the video I describe this as partial automation, because files and business processes can change. A new column, a different date representation or a revised status definition still requires attention. If the process breaks, I want to understand which step failed and why. A refresh button tells me very little about the meaning of the result.
For a practice task, I would retain the first correct result as a reference. After adding the next period, I would check whether the expected orders appear and whether a new filter has unexpectedly removed older records. That habit remains useful when I later work with a larger data process.
I would start with one recurring report whose rules I can explain. This teaches me how to maintain a working solution after its first successful run. Automating everything at once would make it much harder to understand where a discrepancy came from.
Excel belongs in a wider set of tools
I do not expect my learning to end with spreadsheets. The source material also discusses databases, SQL, Power BI and Python. I would choose the next step according to the role I want and the problem I need to solve.
That is an important boundary to my recommendation. If a target role mainly involves databases, I would not postpone SQL until I had explored every Excel feature. If a task is small and happens once, I would not immediately build an elaborate reporting system. I want to judge the appropriate scale of the solution.
AI belongs in that set of tools too. I still keep my own understanding of the data. Questions about keys, dates and missing records remain useful across applications. I discuss this wider perspective in my article about context and analytical skills when working with AI.
For the next exercise, I choose a result I can explain
If technology feels as though it is moving away from you, begin with a limited task. Make a small fictional order dataset, define what each row represents and build a report by category. Trace one result from the source to the summary.
Then change one thing. Remove a category from the product list or add an order. Can you explain what happened to the report? Ask AI for a hint where you get stuck, then compare its proposal with the data and the purpose of the exercise.
Finally, write down what you can now do independently and where you need another attempt. That tells me more about progress than the number of lessons I have watched. Excel still makes sense to me as a place for this concrete practice. Once the foundations work, I add the next tool.
Key takeaways
Does AI make learning Excel pointless?
I still value Excel for quick work with data and conversations with report users. AI can help, but I need enough understanding to judge the result and make ordinary changes myself.
Which skills would I learn first?
I would start with a structured table, filtering, basic calculations, matching records by a key, and PivotTables. Then I would add repeatable data preparation with Power Query.
Should I master all of Excel before learning SQL?
I would not wait until I knew every function. Once I can build a small report, explain its calculations and investigate an error, I can add another tool according to the work I want to do.
AI was used in an automated way while working on this article. If you want to learn how to do it too, check out this project.
