At Systeum Hub: AI Insights, we welcomed Tereza Fukátková from the Data Insights team, who deals with transforming data into values understandable for business. The main question of the discussion was: How is AI changing data analytics and why can everyone work with data, not just specialists?
"30% of code at Microsoft is already generated by AI today. And this is just the beginning," Tereza surprised many of us.
"No code DOES NOT EQUAL no brain." When working with AI, you still need to use your own common sense.
"Journalists who were nominated for the Pulitzer Prize this year almost all use AI in their work." Not in the sense that they would have articles written for them, but for example, they had analyses done that they then used in their articles. (E.g., analysis of E. Musk's behavior based on his tweets). These journalists are a nice example of how data analysis is appearing in professions where it wasn't used before, thanks to AI.
Tereza showed us how not only work with data is changing, but entire professions. Data analysis is appearing in fields where we might not have expected it before, thanks to AI. So let's take a look under the hood.
Revolution in approach to data
"When our Data Insights team was formed, management gave us a clear assignment: help us get value from the data we have," Tereza recalls. It quickly became apparent that the greatest benefit lies in building machine learning models and using generative AI on processed data. This shift from basic business intelligence reporting to advanced analysis represents a fundamental change in how companies approach data. It's no longer just about visualization and simple reports, but about predictive analyses and automated insights.
Interestingly, data analysis is expanding into fields where it wasn't used before, thanks to AI. A BCG study showed significant impact of AI on various professions, including analysts and developers, but surprisingly also on seemingly distant fields, such as the already mentioned journalists.
Jevons paradox: Data analysts can't keep up
Tereza compares the rise of AI to the industrial revolution and the creation of the steam engine, explaining an interesting economic phenomenon known as Jevons paradox. "When the steam engine became more efficient and needed less coal, you would expect coal consumption to decrease. But it was the opposite, it became more accessible, more companies started buying it, so demand for coal increased." The same thing is happening with data analysis. Companies' demand to be "data-driven" is growing, but data analysts can't handle it alone. Therefore, it's up to people in other professions to incorporate data processing into their workflow, which they can suddenly do thanks to AI.
Transformation of roles and new possibilities
Data analysts don't need to worry about their jobs, but their role is transforming in two main directions. First, they focus on more advanced tasks and can do analyses they couldn't handle before. Second, they focus more on data preparation and quality, which is crucial for the success of any analytical project.
Study results are clear: AI significantly helps beginners and speeds up work for experienced professionals. If you're starting with data, AI is a bigger helper for you than for experienced data analysts. AI will speed up work for beginners and help improve quality, for more experienced ones it will only speed up, quality will rather not change.
In practice, you can use AI in 3 main areas:
- Personal assistant
AI helps with writing texts, emails and basic analyses. It can function as a personal analyst for people from various fields.
- Process automation
For example, automatic contract checking in purchasing departments or processing large volumes of documents.
- Part of products
Chatbots on websites, recommendation systems on e-shops or personalized services.
Let's look further at how we can practically use AI as a personal assistant in various professions.
Practical use of AI and its limits
The breakthrough for working with data was adding the code interpreter (advanced data analysis) function in ChatGPT. This function opened the possibility to use generative AI tools for data analysis as well. From uploaded data, it can generate text summaries, graphs, maps, which can even be interactive, and also export results to Excel. The advantage is programming flexibility compared to "boxed" solutions, but there are also significant limitations.
Tools have their limits. They cannot process unlimited amounts of data due to token limits. If you want to analyze data in the order of hundreds of thousands of rows, you'll hit technical barriers. In such cases, however, you can use AI as a programming assistant that will write scripts in Python for processing large data, advise how to divide data into smaller parts, or help with code for analysis.
You won't be alone in using AI as a personal assistant in this sense.
According to data that Tereza presented:
- 30% of Microsoft's code (April 2025),
- 25% of Google's code (October 2024),
- 50% of Meta's code (estimate 2026),
...is generated by AI.
For programmers too, their work is changing significantly, but they don't need to fear being replaced. When working with AI, it's essential to maintain critical thinking. "AI is not a calculator," Tereza warns. "It's a language model, not a mathematical one. If you give it a math problem, make sure it ran the code. If the code doesn't run, it will search for the answer on the internet, which won't be a relevant result." ChatGPT always answers something, never says it doesn't know. Therefore, it's important to still use your own common sense. "No code definitely does not equal no brain."
Use of AI assistants in data analysis
Data analysts in Tereza's team use AI assistants primarily when writing code, which they create most often in notebooks on the Databricks platform with an integrated AI assistant allowing immediate execution of created code. GitHub Copilot works similarly, or Copilot for Excel, which helps with functions and pivot tables, with greater benefit for beginners rather than advanced users.
Tereza emphasizes an important difference between currently popular generative AI tools and traditional machine learning, which is widely used in companies for data processing and predicting future trends. While generative AI can generate texts and images with lesser demands on input data quality, classic machine learning used for predictions and data processing is extremely demanding on data quality and cleanliness. Bad data leads to bad results and requires more careful preparation and more effort in development.
One of the tools that falls into the Machine Learning category and makes data analysts' work easier is Azure AutoML, which, as Tereza also emphasizes, requires really very high-quality data.
Generative AI enabled the team to realize projects they couldn't handle without it
- Sentiment analysis and key topics
They analyzed approximately 15,000 comments (responses to open questions) in various languages.
- Automation of basic insights
Every morning, managers receive 3 A4 tables as a report, which they don't have time to read. The data team's goal was to create one digestible paragraph for managers, highlighting the most essential information.
Getting findings from data analysis to business leaders in an understandable form is clearly a challenge that storytelling helps analysts with. Communication between these two teams needs to start with a clear message: "what I found in the data and what it tells you, what you should do."
And from text, it's just a small step to making it a morning podcast or video. This too is the future of analytics.
Security and 3 concluding thoughts
Not all AI tools are equally secure. While some models (for example, Chinese DeepSeek) don't guarantee data security, there are solutions with well-handled data protection. Tereza recommends models from OpenAI: "paid ChatGPT Team, Enterprise or ChatGPT through API guarantee that input and output data are yours. Even during communication, data are encrypted and cannot be intercepted."
In conclusion...
- AI in analytics is already manifesting now, from BI to AI
- Data analytics isn't just for analysts, people from other fields can use it too and use data for decision-making, not just rely on intuition.
- AI is still a tool that you have to tell what to do.
During the hub, Tereza also received several interesting questions from the audience, see here what was discussed.
Don't miss the panel discussion with all three hostesses either.
If you want to be among the first to know about our future events, subscribe to our newsletter and follow us on social media. Thanks to all participants and their great questions in the live discussion and we look forward to seeing you soon at another Systeum Hub!
🟡 Looking for an interesting project? Check out how things work here and which colleagues we are currently seeking.
🟡 Do you have a colleague or friend who is looking for a new project? Join our Referral program and get a financial reward for your referral.
