While browsing YouTube recently, I stumbled across a channel belonging to Heinrich Rusche. If that name doesn’t ring a bell, here is a quick background: the guy spent nearly six years at McKinsey, went on to work as a CTO for a German retail chain with around 60 stores, and later served as CRO and COO in a scaling tech business. In short: he has extensive experience in core business operations and strategic consulting. Heinrich posted a video where he asked a very provocative yet timely question: in the age of artificial intelligence, do we even need consulting skills anymore?
As I listened to his argument, an immediate lightbulb went off in my head. Why? Because consulting is facing the exact same challenge and superficial narrative that data analysts, BI developers, and software engineers have been confronting for months. From the outside, it looks as though AI could reduce all these professions to zero. Just type a prompt into Claude to generate a presentation, write a Python script, or spit out a complex SQL query, and job done, right?
However, when you examine this closely—from the perspective of someone who has spent over seven years in data analysis and data architecture—it turns out that this grand promise of instant automation is wildly exaggerated.
In this article, I want to break this topic down step by step. We will discuss why reducing expert work down to pure artifacts is a dangerous mental shortcut, why we are currently facing a massive overproduction of worthless AI-generated fast food, and where the irreplaceable human value truly lies in the business decision-making process.
Chapter 1: The Artifact Trap and the Inflation of AI-Generated Fast Food
For years, a certain joke about consultants has circulated in the corporate world. It was said that the only thing consultants are truly good at is creating pretty PowerPoint slides and elegant documents for which companies pay millions, even though they lack deeper substance. A similar accusation is sometimes leveled at analysts: “After all, your job is just writing a SQL query and clicking together a nice chart.”
This brings us to the first widespread mistake in thinking about artificial intelligence, which I would call the reduction of work to an artifact. The logic of surface-level automation enthusiasts looks like this:
- AI can create a document, a presentation, or generate Python code.
- A consultant or analyst delivers a document, a presentation, or code.
- Conclusion: consultants and analysts are obsolete.
This is a massive shortcut in thinking. What we are seeing today in many organizations after becoming intoxicated with text and code generators is not the automation of valuable work, but a massive inflation of sloppy documents. In the industry, people are starting to call it what it is: AI slop. It is akin to fast food—you get a huge dose of calories (words, paragraphs, bullet points), but there is zero nutritional value. Five minutes after reading it, you are hungry for real substance again, and nothing actually comes of it.
In the past, the way to trick the system in a corporate job was creating the appearance of being busy. You walked quickly down the hallway with an open laptop under your arm and a thoughtful expression, and people thought: “Oh, that person has a lot on their plate, they must be a valuable employee.” Today, the new way to fake work is generating tons of text with a single prompt. You get a document packed with specific, buzzword-heavy language where every sentence sounds templated (the classic “It is not just X, it is also Y”), and the bullet points repeat the same platitudes.
Why won’t a real consultant or data architect be replaced by this? Because artificial intelligence takes zero responsibility for its words. We all know those situations with Claude or ChatGPT where, after pointing out an error, the model endearing replies: “You are absolutely right, I apologize, I completely ignored your instructions.” AI suffers no consequences. You, as an analyst or consultant, most certainly do.
I remember from my years working as a Data Architect when we prepared documentation and analytics roadmap plans for the board. Every single word in an email, every footnote on a slide, and every metric defined in a report was subject to fierce debate with managers. Why? Because budgets, team commitments, and company strategy for the coming quarters depended on those words. If you deliver a bad report or form a flawed hypothesis, you will bear the business consequences for months. AI will simply mindlessly accept another prompt.
Chapter 2: Analytical Thinking in Messy Environments (Messy Environments, Messy Data)
Most people looking for a job in data analytics put worn-out phrases on their CVs: “analytical thinking”, “problem solving”. Unfortunately, in practice, this often boils down to a belief that one is technically proficient or simply determined. Meanwhile, real analytical thinking manifests itself under very different conditions than those we know from polished YouTube tutorials.
Heinrich Rusche pointed out a concept in his video that resonates deeply with me: very messy environments, very messy data. And that is the honest truth about working in business.
The decisions we have to make as analysts, BI developers, or architects almost never occur under ideal conditions. Most often, we have to act when:
- There is far too little data, or what exists is of terrible quality.
- Databases are not connected to each other, and documentation is chaotic.
- There is immense uncertainty and conflicting business expectations surrounding the project.
When data is abundant, cleaned, and organized, decisions often make themselves. You just look at the chart. The real art lies in the ability to navigate situations where data is missing, yet the business still has to make a decision: do we implement this tool, do we choose this model version, how do we define this sales metric?
In those moments, the human being gains the upper hand—their experience, their journey, and the number of dirty edge cases they have analyzed. That is why I always tell people wanting to enter the field: do not build a portfolio consisting solely of ideal, beautifully colored reports based on clean, ready-made Kaggle datasets.
If you want to gain real skills, create a small technical nightmare for yourself. Install a database locally on your computer, run into setup errors, let formulas break, and deal with CSV files that have broken character encoding. Struggle through it. It is precisely in those struggles that genuine analytical skills are forged. Solving ten technical problems on your own machine will give you more than watching a hundred perfect courses.
That exact approach guided me when creating KajoDataSpace. I wanted to build a place where you don’t just learn SQL or Python syntax, but interact with real-world problems, learn to build a portfolio resilient to business chaos, and gain mentor support to guide you through those tougher moments.
Chapter 3: Debunking the “Unemployable” Myth and a Sober Look at AI Readiness
Further along in his video, Heinrich makes a very strong and dramatic claim. He asserts that if an employee does not master the ability to build AI-driven processes (meaning combining LLMs with workflow tools), they will become unemployable within a year or two.
Here I have to say firmly: hold your horses, Heinrich, you got a bit carried away.
This is a classic example of alarmist drama that is easy to fall into when living deep inside a tech bubble. If you follow news on X daily, attend every niche event for developers, and listen to speeches by American start-up CEOs, you might fall into the illusion that the entire world around you is already running on autonomous agents.
Meanwhile, reality in the average company looks completely different. When you step into a mid-sized enterprise, you find that:
- Four people in the department have a paid ChatGPT Pro subscription and occasionally paste emails there for proofreading.
- The company has deployed Copilot, which runs slower and produces worse results than an intermediate Excel user (and the more mindlessly employees use it, the worse they become at Excel itself).
- Most processes still rely on Excel files passed around via email.
You will not become unemployed overnight simply because you cannot design a complex, autonomous workflow using LLMs. However, it is worth understanding why implementing artificial intelligence at an organizational level is so difficult.
Privately, you can easily set up a simple automation—for instance, feeding your previous posts into Claude and asking it to write a LinkedIn post in your style. Does it work? It works great. But trying to translate that to an organizational level is a completely different ballgame.
In a company, you run into issues of scale, permission management, data security, API costs, and result repeatability. I always like to use this metaphor: if you enlarge an ant a hundred times, you don’t get a very big ant. You get an elephant. And an elephant is a completely different animal with entirely different anatomy and challenges. Implementing AI in business processes is precisely about turning an ant into an elephant.
Chapter 4: Change Management – Why Technology is Only 10% of the Equation
Anyone who has ever taken part in a major IT project—whether implementing an ERP system, a CRM, or building a data warehouse—knows one fundamental truth. Choosing the architecture or the tool is merely a fraction of the battle. The real uphill climb begins with operational implementation and change management.
Consider CRM systems in massive corporations. Organizations spend millions of dollars on licenses and millions more on consultants. And what is the return on that investment? Often, the adoption rate of the new system among employees is tragically low. People simply do not want to change their habits. They prefer keeping their own notes on the side because the new system requires extra effort from them.
Now multiply that problem by the specific nature of artificial intelligence. Traditional IT systems are deterministic—for the exact same input, you always get the exact same output. LLMs, on the other hand, are non-deterministic systems. Introducing such a tool into daily workflows requires immense flexibility, verification skills, and continuous quality control from employees.
People at the forefront of the tech race often forget how long the tail is of organizations and workers who resist changing the way they work. And that resistance doesn’t stem solely from laziness. It stems from the fact that mindlessly implementing immature technology generates more chaos than value.
That is why the future belongs to individuals who can guide companies through this process. Specialists will be needed who can map business processes, identify real use cases for AI, and then bridge technology with human habits.
Chapter 5: Dashboard is a Noun. Your Job is a Verb
To wrap up, I want to leave you with a thought that is the absolute key for me in understanding the role of a modern data analyst.
I frequently encounter an attitude where an analyst believes the goal of their job is delivering a finished report or dashboard in Power BI or Tableau. “I got the requirements, pulled the data, created the chart, sent the link—job done.”
This is the exact same error Heinrich mentioned regarding consulting. It is equating work with the artifact. A finished dashboard represents a mere 10% of the total value, and by no means the most important part.
Your output is not a static product in the form of a report. Your output is a process.
- Not a dashboard (noun), but reporting (verb).
- Not SQL code (noun), but ensuring data consistency (verb).
- Not a presentation (noun), but driving the right business decision (verb).
Reporting is a set of continuous actions: understanding what and why we measure, identifying source systems, connecting data points, ensuring quality, and enabling decision-makers to actually use it. AI can generate a chart template or write a basic query. But AI will not walk into the Sales Director’s office, ask tough questions about business logic, or spot that a conversion rate spiked simply because someone changed an event definition in Google Analytics.
The future of data analysis is not about fighting artificial intelligence. The future is working with AI, through AI, and because of AI.
Imagine two analytical teams in competing firms.
- Team A pursues total automation: cuts five roles, leaves only artificial intelligence, and expects the system to handle analytics on its own.
- Team B keeps its five analysts, but equips them with AI tools, allowing them to automate tedious, repetitive tasks (like writing documentation or cleaning code).
Who wins that race? Team B wins every single time. Because combining human domain knowledge, business context understanding, and critical thinking with the scale and speed of AI yields an unbeatable competitive advantage.
Summary: Don’t Delegate Your Thinking to Algorithms
Both consulting and data analytics will withstand the wave of automation for one simple reason: business demands truth and accountability, not just fast text generation.
The greatest threat to a modern analyst is not that AI will take their job. The greatest threat is abdicating their own thinking. When you stop analyzing every small problem and start mindlessly pasting queries into ChatGPT, you slowly lose your analytical intuition. It is like multiplication tables—if you need a calculator to multiply 6 by 7, your mental sharpness degrades over time.
Nurture your fundamentals. Learn to understand business processes, develop the skill of asking the right questions, and don’t fear messy data. And if you feel like organizing your knowledge and spreading your wings in this new, hybrid world of data and AI, remember that plenty of practical resources await you at KajoData and within KajoDataSpace.
If this article gave you food for thought and you think it could help someone in your network, please share it on your social media channels (LinkedIn, Facebook, or Twitter). Your shares help me reach more people who want to build a smart, future-proof career in the world of data. Talk to you in the next post!
The article was written by Kajo Rudziński – analytical data architect, recognized expert in data analysis, creator of KajoData and polish community for analysts KajoDataSpace.
That’s all on this topic. Analyze in peace!
Did you like this article 🙂?
Share it on Social Media 📱
>>> You can share it on LinkedIn and show that you learn something new every day.
>>> You can throw it on Facebook – and perhaps help a friend of yours who is looking for this.
>>> And remember to bookmark this page, you never know if it won’t come handy in in the future.
You prefer to watch 📺 – no problem
>>> Subscribe and watch my English channel on YouTube.
Prefer to read in Polish? No problem!
Other interesting articles:
