I’m a software engineer who became part of the data industry back in 2013. Since then, I’ve been working as a data engineer and architect, which means I’ve spent an unreasonable amount of my life moving data from one place to another and explaining to people why their reports don’t work. I’ve watched Power BI grow from its early days, and I’ve got opinions. Strong ones. Especially about pivot tables.
I’ve been trying out the latest Standalone Copilot release, and it got me thinking about how far we’ve actually come. Not just with this one feature, but with the whole idea of asking questions in plain English instead of building pivot tables for everything. Turns out there’s quite a story there, and I thought it was worth writing down.
Right, let’s talk about pivot tables. Specifically, let’s talk about how much time we’ve all wasted on them.
Not because pivot tables are bad, they’re not. They’re brilliant, actually. Excel’s pivot tables are one of the most useful features ever created. But here’s the thing: they’ve held us hostage for decades. Every simple question, every straightforward bit of information, every “just show me the top five” request required building a pivot table, dragging fields around, filtering, sorting, reformatting, and hoping you hadn’t accidentally created a subtotal where you didn’t want one.
And we all just… accepted this. Like it was normal. Like spending ten minutes setting up a table to answer a thirty-second question was a reasonable use of anyone’s time.
Well, it bloody well wasn’t. And the ten-year journey of natural language in Power BI is the story of how we’re finally escaping that tyranny.
The Pivot Table Prison
Let me paint you a picture. It’s 2013. Someone walks up to your desk and asks: “What were our top five products last quarter?”
Simple question, right? Straightforward information. Should take seconds.
Here’s what actually happened:
You open the report. Find the right data table. Create a pivot table. Drag “Product” into Rows. Drag “Sales” into Values. Hope it’s summing and not counting. Filter to last quarter, but wait, which field is that? Is it “Order Date” or “Ship Date” or “Fiscal Quarter”? Try one. Wrong. Try another. Still wrong. Finally find the right one.
Now sort by sales descending. Take the top five. Format the numbers so they don’t look stupid. Remove the grand total because it’s not useful here. Maybe add a percentage if you’re feeling fancy.
Ten minutes later, you’ve got your answer. For a question that should have taken five seconds.
And if they then ask “what about the quarter before that?” – well, you’re starting over, aren’t you?
This was just how things worked. We were so used to it that we didn’t even question it. The data was there. We had tools to access it. Those tools just required a specific set of skills and a lot of patience.
The First Attempt: Q&A in 2014
When Microsoft launched Q&A in Power BI in August 2014, the promise was simple: Just ask questions in plain English. No more pivot tables. No more dragging fields. Just type “top five products last quarter” and get an answer.
It sounded too good to be true. And it was.
I tried it. We all tried it. You’d type a question, and one of three things would happen:
- It would completely misunderstand you
- It would sort-of understand you but create the wrong visual
- It would work perfectly, and you’d think “brilliant!” – then the next question would fail spectacularly
The problem was that Q&A expected you to ask questions in a very specific way. It needed you to use exact field names from your model. If your column was called “SalesAmount” but you typed “sales”, it might work or might not. If you had similar fields with different meanings, it would guess. Usually wrong.
And here’s the problem if your data model wasn’t pristine. Which, let’s be honest, whose data model was pristine in 2014? We were all still working out how this Power BI thing worked.
So what did we do? We went back to pivot tables. Because at least with pivot tables, you knew what you were getting. It might take ten minutes, but it would work. Q&A might work instantly or might waste fifteen minutes of your life before you gave up and built a pivot table anyway.
Better the devil you know, right?
The Years Nobody Talks About: 2015-2018
Q&A didn’t die. Microsoft kept updating it, adding features, improving the language engine. But nobody cared. Or more accurately, nobody used it enough to care.
I remember this period clearly. I was building data solutions, creating Power BI reports for clients, training analysts. And Q&A never came up. People wanted to know about DAX, about modeling, about performance. They wanted to understand relationships and how to create measures that didn’t break when you filtered by date.
Nobody asked about Q&A because everyone had tried it once, found it unreliable, and moved on.
But here’s what was actually happening during those quiet years: Microsoft was learning. Every question someone typed into Q&A, every failure, every time someone gave up and went back to manual methods – all of that was data. Data about what people actually needed versus what the system could do.
This is the bit that’s easy to miss. We think of improvements as sudden breakthroughs. They’re not. They’re years of incremental fixes. Someone notices that people often type “sales” when the field is “SalesAmount”, so they add fuzzy matching. Someone realises that questions with dates are failing because of format issues, so they improve date parsing. Someone sees that people want top-N results but the system is showing everything, so they add ranking logic.
Hundreds of small improvements. Each one making the system slightly less terrible. None of them enough on their own to change anyone’s mind about using it.
Meanwhile, we were all still creating pivot tables.
2019: Something Actually Changes
The redesigned Q&A visual launched in 2019, and something was different this time.
First off, it was actually a visual now. Not a separate thing living on dashboards somewhere. A proper visual you could add to your reports. This sounds trivial, but it wasn’t. Because now Q&A wasn’t competing with your entire workflow, it was part of it. You’d build your standard visuals as always, then add a Q&A box for ad-hoc exploration.
But more importantly, it actually worked reliably enough that you could trust it.
The key was feedback. As you typed, it would show you which bits it understood. Fields it recognised would light up. It would suggest completions. If it was confused, it would tell you. If your question was ambiguous, it would offer options.
This completely changed the psychology. Instead of typing into a void and hoping, you were collaborating with a system that was trying to help. You could see what it understood and adjust your question accordingly.
And it was fast. Properly fast. Not “technically fast if you consider the complexity of the language processing” fast. Fast like “I’ve got my answer before I’ve finished thinking about whether to create a pivot table” fast.
For the first time, asking a simple question in natural language was actually easier than building a pivot table. Not just potentially easier. Actually, genuinely easier.
The Accidental Accelerator: Working from Home
Then 2020 happened and we all went home. Video calls became meetings. Screen sharing became normal. And suddenly Q&A had a new use case nobody had anticipated.
Live collaboration.
You’re in a video call. Someone asks a question. Pre-2020, you’d say “let me pull up that report” and everyone would wait while you filtered things and created visuals. Or more likely, you’d say “I’ll get back to you on that” and add it to your ever-growing list.
With Q&A, you just type the question and share your screen. The answer appears instantly. The meeting keeps flowing. People ask follow-up questions, you type them in, more answers appear. The whole conversation stays focused instead of fragmenting into “I’ll come back to you” promises.
This wasn’t what Q&A was designed for. But it turned out to be exactly what remote collaboration needed. We still needed proper reports for the important stuff. But for those ad-hoc questions that come up when people are discussing data? Perfect.
The pivot table alternative had found its moment. Not by replacing everything. By being better for specific situations.
2025: Copilot and the Final Freedom
Which brings us to Copilot. The Standalone Copilot experience that rolled out in 2025 isn’t just “Q&A version 2”. It’s what Q&A should have been all along if the technology had been ready.
We’re now in the era of generative AI, and it shows. Power BI already had Quick Insights, a feature that would automatically spot patterns and trends in your data. You’d click a button and it would show you things like “sales went up in March” or “this product is an outlier”. Useful, but limited. It could only tell you what it was programmed to look for.
Copilot is different. It’s not just finding patterns, it’s having a proper conversation about your data.
Remember how Q&A only worked on the current report? Copilot works across everything. All your reports, all your models, all your data. You don’t need to know which report contains the answer. You just ask the question.
Remember how you had to phrase things carefully? Copilot understands context. It remembers what you asked before. You can have an actual conversation. Ask “top products”, then ask “what about last year?” and it knows you mean top products last year. No starting over. No re-explaining.
Remember how it would fail on complex questions? Copilot can break down complex requests. If it can’t answer directly, it’ll suggest alternatives or ask for clarification.
And here’s the crucial bit for getting away from pivot tables: It’s not just showing you data. It’s explaining it. Creating summaries. Spotting patterns. Doing the analysis you’d normally do after creating the pivot table.
This is the real freedom. Not just getting data faster, but getting insights faster. The thinking that used to happen after you’d built your pivot table now happens automatically.
What We’ve Actually Escaped
Right, let’s be specific about what’s changed. Because “liberation from pivot tables” sounds dramatic, but what does it actually mean day-to-day?
The Ten-Minute Question
You know the ones. Someone asks something simple. You know the answer is in the data. Getting to it requires setting up a table, filtering, sorting, maybe creating a quick calculation. Ten minutes minimum.
Now? Type the question. Get the answer. Thirty seconds. That’s not hyperbole, that’s literally what happens.
Multiply that by the number of ad-hoc questions you get each week, and it could add up to a significant amount of time saved.
The “I Don’t Know Where That Is” Problem
Pre-Copilot, when someone asked a question, you had to know which report contained the answer. Or which model. Or you’d go hunting through workspaces trying to remember where you put that data six months ago.
Now you just ask. Copilot finds the right source. If there are multiple options, it shows them and lets you pick. You don’t need to be the keeper of the data catalogue anymore.
The “Let Me Get Back to You” Tax
This was the killer. Someone asks a question during a meeting. You don’t have time to build a pivot table there and then. So you make a note, do it later, send them the answer. Except by then the conversation has moved on and the insight isn’t as valuable anymore.
Now the answer happens during the meeting. While people care. When it can actually influence the decision. The value isn’t just in getting the answer, it’s in getting it when it matters.
The “I Need Someone Technical” Bottleneck
This one’s subtle but huge. When getting straightforward information requires knowing how to build pivot tables or filter Power BI reports, only certain people can access data. Everyone else has to ask. Which means analysts become bottlenecks.
Natural language doesn’t eliminate that completely. Complex stuff still needs expertise. But for simple questions? Business users can help themselves now. They can check their assumptions. Verify their hunches. Get on with their work.
That’s not about technology. That’s about changing who has access to information.
What Hasn’t Changed (And That’s Fine)
Let’s not get carried away. We haven’t eliminated pivot tables. We haven’t made data analysis trivial. We haven’t solved all problems with natural language queries.
Complex analysis still needs proper tools. If you’re building production reports, doing sophisticated calculations, or creating dashboards for executive review, you’re still using traditional methods. Copilot is brilliant for exploration. It’s not replacing structured analysis.
Data quality still matters massively. Actually, it matters more. Bad data plus AI just means wrong answers delivered with confidence. If your model is rubbish, Copilot will struggle. Garbage in, garbage out hasn’t been repealed.
You still need to know what questions to ask. This is the bit people miss. Natural language makes it easier to get answers. It doesn’t tell you what questions matter. That still requires business knowledge, experience, and judgement.
Some things are better with pivot tables. If I’m doing proper exploratory data analysis, if I want to play with different arrangements, if I need precise control over grouping and aggregation – I’m probably still building a table manually. The control is valuable.
What we’ve escaped isn’t pivot tables themselves. It’s the tyranny of needing them for every trivial question.
The Psychology of Liberation
Here’s what I didn’t expect: The biggest change isn’t technical. It’s psychological.
When getting an answer requires ten minutes of setup, you only ask important questions. You don’t explore hunches. You don’t check assumptions. You don’t follow curiosities. Because the overhead is too high.
When getting an answer takes thirty seconds, you ask different questions. Smaller questions. Weirder questions. Questions you’d never have bothered with before because they weren’t worth the time.
And sometimes, not often, those throwaway questions lead somewhere interesting.
I’ve seen this happen repeatedly. Someone asks something casual, gets an unexpected answer, asks a follow-up, discovers something nobody knew. None of that would have happened if each question required building a pivot table.
The liberation isn’t just about time. It’s about changing what questions feel worth asking.
There’s also something about confidence. When business users can verify things themselves, they trust the data more. When they can dig into the “why” without waiting for someone technical, they feel more ownership. When data stops requiring a specialist to access, it becomes part of how they work.
That’s harder to measure than time saved. But it might be more valuable.
A Decade Well Spent
So here we are. Ten years from Q&A’s launch in 2014 to Copilot in 2025. A decade of iteration, improvement, and quietly fixing things until they worked.
It didn’t happen through some brilliant breakthrough. It happened through Microsoft watching people struggle, understanding what wasn’t working, and improving it. Then improving it again. Then improving it some more.
Was it worth the wait? Yes, definitely. Because what we’ve got now actually solves the problem. Sometimes not perfectly, but well enough that it’s changed how I work.
I still build proper reports. I still write DAX and optimise models and do all the traditional data work. But for those simple questions, those ad-hoc requests, those “just quickly show me” moments that used to eat up time? I just ask. Get my answer. Move on.
That’s the liberation. Not from pivot tables entirely. From needing them for everything.
The Future (Probably)
Where does this go next? I haven’t a clue, to be honest. But I can make some guesses based on where we’ve been.
More contextual awareness. Copilot will get better at understanding not just what you asked, but what you probably meant. Your role, your patterns, your organisation’s quirks.
More integration with actual work. You’ll ask questions in Teams, in email, wherever you actually work. The separation between “using Power BI” and “just working” will blur.
Learning from corrections. When you tell it an answer is wrong or not quite right, it’ll learn. Not just generally, but for your specific context.
But honestly? The specifics matter less than the direction. We’re moving away from “you must use specific tools in specific ways to access information” towards “just ask and the system works it out”. That trajectory is clear.
The pivot table prison isn’t being demolished. It’s becoming optional. And that’s good enough.
I started this article talking about natural language like it’s a novelty. It’s not. It’s a quiet shift that’s been building for years.