Are We Automating Chaos with AI Transformation?
Years ago—if I remember correctly, it was around 2014 or 2015—an organization with thousands of employees across both central and regional units contacted me. They wanted to digitize their physical archives, enable access to digital data through software, create and manage new data via a system, and fully integrate with various software products used by different business units. At first, the project seemed reasonable and feasible—until we learned that the number of systems requiring integration was over 100. Thinking we had misunderstood, we asked again. When we did, the responsible manager felt the need to clarify:
“As the technology and software sector evolved, each department either developed or purchased many small applications tailored to its own needs. These systems have been used for years, but now maintenance costs are high and managing them has become extremely difficult. While integrating them, we also want to simplify.”
Whenever I see how prominent AI has become—and how many organizations are trying to “do something” with AI—this moment comes back to mind. As technology and human expectations change, transformation becomes inevitable. Organizations attempt to evolve based on what is necessary at the time. Sometimes they succeed; sometimes they don’t.
Just as the digitization of physical data and the subsequent wave of software adoption defined the past, today we are experiencing a similar wave—this time around AI. In organizations, I often observe attempts to adopt AI in the following perceived sequence:
- Building chatbots
- "AI-supported" products, services, and software
- AI agents
However, this sequence is less an actual evolutionary path and more a market perception.
If we look at a simplified chronological progression, it might look like this:
- Pre-AI (automations like RPA that execute tasks but don’t make decisions)
- Statistical AI (systems that suggest but do not decide)
- NLP & Conversational UI (a common misconception: “we built a chatbot, so we’re using AI”)
- LLMs (entering the realm of intelligence)
- AI-supported systems (copilots; AI begins to support humans)
- Agentic AI (takes goals, plans actions, uses tools, and progresses through feedback)
Not every organization follows—or has followed—this exact path. Experimenting and adapting to new technologies is valuable. However, jumping directly from chatbots to AI agents can lead to disappointment. Launching “AI transformation” solely as an IT-driven project may also result in the risk I mentioned earlier: ending up with 100+ uncontrolled agents.
Transformation does not end with adopting Agentic AI—quite the opposite. Treating it as just another IT initiative is risky. Efforts left at that level may simply automate existing organizational chaos. What’s needed instead is to integrate AI agents into processes to create AI-first value streams, and from there, move toward becoming an Agentic Organization.
So, what is an Agentic Organization?
At its core, it is an organizational approach where the boundaries of who decides what—human or AI—are clearly defined; where not just data flows but value is produced autonomously; and where the organization learns quickly and continuously adapts.
We will see what else AI brings in the future. But in my view, true transformation will not happen when we discover what AI can do—it will happen when we experience what we are willing to let it do.
Comment(s)