Generational conflict or opportunity? How to make the most of the differences between Gen Z and managers.

Companies worldwide are trying to use large language models as an additional analytical tool and are encountering fundamental limitations of the technology. However, these models were not created for spreadsheets. They are designed for language. This means that the greatest strength of today’s AI does not lie in BI warehouses. It lies in how you speak, write, decide, and communicate within the company. In emails, minutes, presentations, and meeting transcripts, there is far more untapped value than in most dashboards. And that is where AI can start working immediately. Without projects, without migrations, without waiting.  

Most companies think that deploying AI first requires order in data and solid foundations. It sounds responsible, but it is a mistake. It does not mean that it is unnecessary to invest in data infrastructure, but that it is nonsense to wait for "real" AI deployment for another year or two, until large data projects are delivered. 

Typical enterprise IT fundamentally still runs on an old mental operating system that has stretched since the 1990s. Big problems. Big solutions. Multi-year roadmaps. Huge budgets. Transforming a company is actually understood as a change in infrastructure, not as a change in behavior. That worked in the cloud era, but it no longer fits the physics of today’s AI quite as much. 

Language models mainly need context 

Large language models do not need structured data. They need context. And businesses are flooded with it. Every larger company today sits on an unmonetized asset. Millions of words. Emails, presentations, documents, Slack threads, call transcripts. The living language of how the company truly operates. 

Instead of waiting for the data to be perfect, feed AI with the real language of the company today and enhance your strategy. Provide AI with unrefined meeting transcripts, presentations, older documents. AI will extract patterns from them and create understandable syntheses that can greatly improve managerial judgment very quickly and refine not only the definition of strategy but also ensure consistency in its execution.  

This turns the logic of AI transformation somewhat upside down. A bottom-up approach (clean data, models, insight, action) is today slower, more expensive, and less effective for established companies than a top-down approach (language, context, judgment). Companies are unnecessarily waiting for data and systems to mature properly. They do not see that LLMs are already unlocking the most valuable "soft" layer of the organization: decision-making, strategy, communication, and customer signals. Everything that is expressed in human language. 

Today’s AI does not model the world 

Why build a language model with perfect data structure only for it to occasionally hallucinate over them anyway?  LLMs are not analytical machines. They are inferential machines over language. They do not model the world; they model how people describe it. Therefore, a company’s true readiness lies not in the data warehouse but in its language. In how people talk and think inside the company. That language already exists, but no one is utilizing it. 

Some time ago, we began to systematically transcribe and analyze all work conversations in our company. We got used to smart meeting minutes that not only summarize conversations but also highlight what we forgot and send people developmental feedback on how they communicate a topic. When I come to a meeting today without recording, it feels as if you are heating with an open window. Words and thoughts are the money that companies today waste out the window.  

Sometimes clients openly tell us: "We are not prepared for such a level of transparency."  Top-down AI indeed takes away organizations' favorite alibi of a lack of IT resources and shows that the barrier to change is actually cultural. When leadership can create the first iteration of the strategy in 10 seconds, the bottleneck is no longer IT. It is the people, their habits, and identity. 

This leads me to a perhaps somewhat bold thought that the winning products of this generation will not be data platforms. They will be cognitive infrastructure systems. Systems that unify language. Reveal drift. Align teams with shared context. Compress ambiguity into clarity. Enhance human judgment at scale. Ensure consistent interpretation of strategic decisions across the organization. This will not be infrastructure, but a new layer of thinking. This will be the next McKinsey. Not as consulting, but as a permanent cognitive layer over the business. 

How does a top-down AI transformation begin?  

Not as a project. 
Not as a roadmap. 
Not as a program. 

It begins with five simple steps: 

1. Start with judgment, not data. AI should always create the first version of anything you need to discuss.  

2. Feed the model real language. Unrefined documents, meeting transcripts, presentations, and dusty folders. 

3. Leadership must lead by example. If AI is not used by leadership every day, no one is using it. 

4. Make AI a habit. No one starts from scratch. 

5. Expand only where there is natural pull and you immediately see value. 

This is the least ceremonial transformation in corporate history. And at the same time, the most strategic.