Picture a regional sales manager on a Monday morning. She asks an AI tool for next quarter’s forecast, and forty seconds later it’s there: a clean chart, numbers to one decimal place, and a short paragraph explaining the trend. She drops it into the leadership deck and moves on.
Nobody asks where the numbers came from. Nobody notices that one region’s figures were in thousands and another’s weren’t.
AI didn’t create that problem. It just made it faster. These tools work from whatever data they’re handed, and they’ll present a broken dataset with the same calm confidence as a clean one. The person who catches the error, who looks at the chart and thinks “hang on, that can’t be right,” is suddenly very valuable. And that person doesn’t have to sit in the analytics team. More and more, it’s the marketer, the ops lead or the finance manager.
Employers already know this
The World Economic Forum’s Future of Jobs Report 2025 still puts analytical thinking at the top of its list of core skills (seven in 10 companies call it essential), and it names AI and big data as the fastest-growing skill of all.
What workers actually do with AI is more worrying. A 2025 study by KPMG and the University of Melbourne, covering more than 48,000 people in 47 countries, found that 58% of employees deliberately use AI at work. Two-thirds of them, 66%, rely on what it produces without checking whether it’s accurate. And 56% say they’ve already made mistakes in their work because of it.
So most people use it, most don’t check it, and over half have been burned. That’s the gap.
The quiet ways it goes wrong
AI almost never fails dramatically. It fails in ways that look completely reasonable until someone who knows the data takes a second look.
The mislabeled column is the classic one, like the sales manager above. A region that didn’t report for a month gets read as a region that sold nothing, and the summary dutifully announces a collapse. Ask for “the key trends” and you may get a confident story about a pattern that’s really just random wobble. Chatbots doing sums directly on spreadsheet data still slip up on arithmetic, too. Not often, but often enough that you wouldn’t want an unchecked number going to the board.
You only catch these things if you understand what’s underneath. Garbage in, garbage out isn’t a new idea. It’s the whole reason companies invest in proper data engineering for AI. But it applies just as much to one person with a spreadsheet and a chatbot as it does to a data platform.
Five minutes of checking
None of this needs a statistics degree. It mostly needs the habit of asking a few awkward questions before an AI-generated analysis goes anywhere important.
Start with the input. What data did the tool actually use, and is it complete, recent and labeled the way you think it is? Then look at the size of the numbers. If a product that’s sold steadily for three years suddenly shows a 400% jump, that’s not a breakthrough. It’s probably a mistake. Pick one figure, any total or average, and work it out yourself. If it doesn’t match, stop there.
Ask what the trend is being compared against, because “up 12%” means nothing without a baseline. Scan for gaps, duplicates and odd outliers, since they’re behind a surprising number of surprising results. And keep a clear line between what the data actually shows and what the AI decided it means. Those are often two very different things.
That’s it. It takes about five minutes, and it’s the cheapest insurance you’ll ever buy.
What not to paste in
Data skills also include knowing when to keep data away from AI. Customer lists, salary sheets and unreleased financials don’t belong in a public chatbot. Doing that can leak sensitive information, and depending on where you work, it may break company policy or data protection law. The KPMG study found that plenty of employees are already using AI in ways their own organizations don’t allow.
It helps to have a rough mental filing system: is this public, internal, confidential or personal? If you’re unsure, use whatever AI tools your company has approved, or strip out names and identifiers before you paste anything.
The translator role
Every business of any size runs on systems that never stop producing data: ERP software, supply chain platforms, CRMs. The trouble is that the people who understand those systems and the people making decisions often barely speak the same language. Decisions stall. Two departments buy the same tool. A strategy drifts quietly away from what the numbers say.
The people who fix that are translators. They know enough business, whether accounting, market research or how a team actually runs, to understand what leadership needs. And they know enough about data to turn that into clear requirements for the technical side. They can also walk back into the boardroom and explain, in plain money terms, why the thing everyone wants isn’t possible yet. It’s not a job title. It’s a reputation, and it tends to lead to bigger roles.
Getting the skills
The old choice between a business degree and a computer science degree has mostly gone away. Plenty of universities now blend the two.
A business degree with a serious analytics component is one way in. Someone who wants to go further into the technical side, perhaps heading toward a technology management role, might look at an online information technology degree instead. IT programs center on managing and supporting technology systems, networks, and databases, and many include cybersecurity tracks. The online format is the practical part: you can keep your job while you study. People who already have a degree sometimes go for an MBA with an analytics focus as a step toward senior management. It really comes down to where you want to end up, and how much time and money you can spare.
If you’re weighing online programs, the curriculum matters more than the marketing. Look for real, hands-on work with the tools you’d use in a job, plus case studies and projects built around messy, realistic business problems rather than textbook exercises. Check that the school holds accreditation from a recognized body. And if you can, talk to a couple of graduates about what the degree actually did for them.
The tools, and the AI now inside them
Whatever route you take, a few tools come up again and again in business roles, and almost all of them have AI built in now:
Tool | What it’s for | The AI layer |
| SQL | Pulling, filtering, and summarizing data from databases | AI can draft queries, but you still need to check the logic |
| Excel | Budgets, forecasts, pivot tables, everyday analysis | Copilot in Excel suggests formulas, analyzes data, and works with Python in Excel |
| Power BI / Tableau | Dashboards and reports for leadership | Copilot in Power BI, plus Tableau Agent and Tableau Pulse, turn plain-language requests into visuals and insights |
| Python or R | Automating data cleanup, statistics, simple models | AI assistants can write the code, but understanding it is what lets you trust it |
The AI makes each of these quicker and less intimidating. What it doesn’t do is make understanding optional. If Copilot writes you a formula you can’t read, you’ve no way of knowing whether it’s right. You’re back to trusting the confident chart.
Where this leaves you
Routine reports and admin are exactly what AI is eating first, and it’s natural to feel uneasy about that. A lot of people share the same fear of becoming obsolete. But the people who’ll do well aren’t the ones avoiding AI. They’re the ones who can look at what it produces, tell good output from bad, and explain the difference to everyone else in the room.
Add decent communication and some leadership on top of that, and you’re not in competition with the software. You’re the reason it’s worth having.
Frequently asked questions
Q. If AI can analyze data, why do I still need data skills?
Because AI only knows what it’s given. Without some data know-how, you won’t notice when a slick answer rests on bad inputs, missing values or a wrong assumption.
Q. Which skills should I start with?
Advanced Excel and basic SQL come first for most people. Then add a dashboard tool like Power BI or Tableau. Python or R is worth learning after that if you want to automate work or run proper statistics.
Q. How can I tell if an AI analysis is accurate?
Check the source data, see whether the numbers are plausible, recalculate at least one figure by hand, and keep the observed facts separate from the AI’s interpretation of them.
Q. Can I paste company data into ChatGPT?
Be careful. Keep confidential and personal information out of public AI tools. Use your company’s approved tools, or remove identifying details first.
Related: Human Skills in the AI Era: 6 Skills That Matter Most in 2026
