A transformation is a ribbon of decisions, taken in small and in large matters. Whether these decisions are made deliberately and aligned with each other, or whether they trigger a certain behavior without leadership, makes a great difference.
The same technology can thus lead in very different directions, depending on whether leadership sets the course. None of the patterns on the left is a technology problem; behind each stands a leadership decision nobody has taken yet.
The table below is meant to prompt you to think about the effect of a curated, clearly steered AI transformation in your company.
What costs time and nervesUnled AI‑transformation
And what does the oppositeLed AI‑transformation
Strategy
Cost and headcount savings in the foreground; employees are afraid. Plenty of activity, no focus on value. Pilots remain pilots: 95% of the GenAI‑initiatives examined showed no measurable effect on results.1
A clear target picture of how humans and AI are to work together and where AI can create the greatest value. A few prioritized use cases at a time — employees see opportunities and prospects. Clear criteria for success and for stopping.
Framework
Shadow AI: many people use private AI for work as well, yet it is hardly discussed on the job. There is no shared framework, and no one knows which data flows where.
Employees have AI support that goes beyond help with wording. Clear rules of play instead of bans, a deliberate time allowance and active enablement for case-based experimentation and exchange.
Quality ofwork
People rely on AI and quickly produce fine-sounding documents that simulate work and create extra work for the recipient — workslop.2 An inflation of documents, applications and concepts that no one can read anymore. Quality of work and focus decline.
A manageable number of topic-specific AI assistants and agents that strengthen employees in a goal-oriented way. AI uses uniform standards; customers experience faster responses and better advice. User reviews and impact measurement are part of every use case.
Data and systems
Three departments, three figures for the same business transaction. The company becomes dependent on external vendors and settles for the standard product. The AI answers with complete conviction, and employees are busy mainly reconciling numbers.
Business-critical terms are defined bindingly. Every agent is trained by experienced experts and accesses the right internal tools according to a consistent logic. Employees learn to lead AI — using AI becomes a matter-of-course skill in daily work.
People and organization
People fear the future and are wary of AI. Customers are fobbed off with AI agents. Responsibility diffuses, the offering becomes interchangeable, the company unattractive as an employer.
People work with AI in a natural way. Agents have a clear identity, their actions are traceable, and there is always a human who leads them and bears the responsibility. People gain time for the situations in which trust or creative ideas emerge.
Which of the two pictures is your company currently extending, and how can you tell?
1 MIT NANDA, The GenAI Divide: State of AI in Business 2025 — analysis of 300 public AI deployments, 150 interviews with executives. 95% of pilot projects without a measurable effect on operating results.
2 BetterUp Labs and Stanford Social Media Lab, Harvard Business Review, September 2025 — survey of 1,150 employees in the US: 40% received workslop in the previous month, with an average of 1 hour 56 minutes of extra work per incident.
Where this approach reaches its limits
The comparison simplifies. Hardly any company sits entirely on the left or entirely on the right; most mix the patterns, depending on the unit and the leader. The table helps you recognize your own mix; it does not measure it.
The right-hand column assumes a management team that wants to decide and is willing to be held to its decisions. Where that is missing, even the best target picture produces just another document. And where regulation, data protection or works agreements set the frame, part of the rules of the game is not freely chosen.
The patterns come from MLI projects and from the two studies cited. They are observations, not statistics about all companies. Anyone who wants to use them as a diagnosis of their own organization needs the survey behind them, not just the comparison.