Excelstentialism in the AI Age: One Day at a Time

It seems like every day brings another debate about what artificial intelligence means for the future of work.

Will AI eliminate entire professions? What happens to the expertise people spent decades developing? Will we need universal basic income? Will anyone still know how to do anything themselves?

These are fair questions. I’m certainly not immune to wondering about them myself.

But lately I’ve been wondering how much of this discussion, particularly at the personal and organizational level, has become rumination rather than reflection that leads somewhere.

At some point, what are we actually supposed to do with all of this?

Your Friday report still needs to get done

I work with organizations trying to modernize how their employees use Excel, data analytics, and increasingly AI. There is plenty to think about: security, accuracy, changing job responsibilities, and whether any of these investments will pay off.

Those questions deserve attention. But you can spend six months discussing what AI might do to the finance department while that same department is still manually consolidating spreadsheets every Friday.

That’s a problem you can actually do something about.

Maybe Power Query could bring those files together. Maybe the workbook just needs a better structure. Or maybe two departments need to agree on what a column actually means before introducing another tool.

I’m not suggesting that every business needs to rush into AI. Sometimes the most useful improvement has nothing to do with AI at all.

What concerns me is when speculation about the future starts crowding out improvements we can make in the present. The discussion takes time, the frustrating process stays the same, and the people doing the work are still waiting for help.

A little Excelstentialism

One of the things I’ve come to appreciate about working with Excel is how incremental the improvement can be.

You don’t have to transform an entire reporting process in a single afternoon. You might start by converting a range into an Excel Table. Tomorrow, perhaps you learn how structured references work. Next week, you use Power Query to eliminate a few repetitive cleanup steps.

You try something, check it, and build on what you learn. Over time, the workbook becomes easier to maintain and the person using it becomes more confident.

None of this requires predicting what spreadsheets will look like in 2035.

Call it Excelstentialism: accepting that you don’t have every answer about the future, while taking responsibility for the work you can improve today.

That’s very much how I think about Stringfest Analytics. I want to help people move from fragile spreadsheets and recurring workarounds to analysis they understand and can rely on. New tools are useful when they help us get there.

What this looks like in practice

In my own work, I’ve been experimenting with Copilot in Excel, Python in Excel, reusable AI skills, and conventional automation. Some experiments work better than others. Sometimes the older approach is still the better one.

I’ve found that smaller, concrete experiments give me something useful to evaluate. Does this make a task easier? Can I check the result? Is it something I could use again or teach someone else?

Here are four ways to put that mentality to work in your own organization.

1. Fix one repetitive reporting task

Find something your team does over and over again: cleaning monthly exports, combining files from several departments, or updating the same report.

Write down where the time goes and what tends to go wrong. Then see whether Excel Tables, Power Query, or a better-designed workbook could simplify the process.

Try the revised process on a copy of the data. Compare its output with a result you trust, including totals and any records that need special handling. Time the work again, including the checks.

If a refreshable Power Query solution does the job, great. You have a useful improvement without needing to build an elaborate AI agent.

2. Try AI on one specific task, then check it

Take a real problem, perhaps understanding an unfamiliar formula, getting suggestions for a report, or drafting an initial analysis of a dataset.

Use a tool approved by your organization, with data you’re permitted to use in it. Compare the output with what you know about the work. Check the formulas, assumptions, and conclusions.

Was it faster once you included the review? Was it accurate? Did it help you understand the problem better? Would you use the approach again?

Those answers can help you decide what to do next. Another prediction about the fate of every analyst in the building probably won’t.

3. Make a successful experiment repeatable

One area I’ve been exploring is reusable instructions and skills for common tasks. If you’ve developed a useful way to review a spreadsheet, explain a variance, or check a report, why reinvent those instructions every time?

Document the inputs, the steps, and the checks that tell you whether the result is usable. Try a few representative examples, including one that doesn’t fit neatly. Make the approach available to a colleague where appropriate.

A successful demonstration is a start. Being able to repeat it, explain it, and recognize when it has gone wrong is what makes it useful in the business.

4. Help one colleague learn something useful

Not every improvement needs to involve a new tool.

Show someone how to use a PivotTable. Explain why a Power Query step works. Walk through how you verified an AI-generated answer, including anything you had to correct.

Give them enough context to try it themselves and judge the result. Building your team’s confidence and judgment can pay off across many different tools.

Take it one day at a time

Of course you need a broader strategy. You need policies, budgets, and some sense of where the organization is heading.

But that strategy should eventually translate into things people can do during an ordinary workday.

If you’re responsible for a team, try this small exercise. Keep the scope narrow enough that you can learn something by Friday.

Some weeks you’ll find a way to save your team time. Other weeks you’ll discover that a promising tool isn’t ready for your needs, or that the existing process works better than you thought.

Either outcome gives you something concrete to build on. You can stop doing the things that don’t help and spend more time on the ones that do.

There’s still plenty to look forward to

One thing I find discouraging about the existential discussion around AI is how easily it can crowd out curiosity.

There are things I can experiment with today that would have required considerably more time and technical expertise just a few years ago. That’s exciting!

It doesn’t mean everything new is better, or that the skills we’ve developed suddenly have no value. Knowing what a report should show, which assumptions need challenging, and when a result doesn’t make sense is part of what makes these tools useful.

I understand the satisfaction of solving a problem yourself. I also understand the concern that we could lose some of that understanding if we delegate too much. Those are good reasons to stay involved in the work.

But I’d hate to become so preoccupied with everything we might lose that I stop exploring what I can do now.

We don’t need every answer about AI, employment, or the economy before making sensible decisions about our own work.

Take it one day at a time. Find something worth improving, try something, check what happens, and keep learning.

That seems like a pretty good way to approach the future, whatever it happens to bring.