O

Most enterprise AI projects do not die because the wrong tool was chosen. They die because the organization switched on artificial intelligence faster than its own structure and culture could process it.

This is the sentence leaders rarely hear — even though it sits behind most failures. The model was not weak. The organization could not handle the pace.

The SICT framework — developed by Miklós Róth (Roth Complexity Lab, Budapest) — shines a light precisely on this blind spot. It is not another “AI maturity score” that hides reality behind a number. It is a diagnostic lens that gives you better questions.

The real source of fragility

SICT is built on a single core thesis, and that thesis is uncomfortably precise:

Modern systems do not become fragile because they have become too complex — but because information and change flow through them faster than structure and cohesion can process.

Now think through what an AI implementation does to a large enterprise. In one stroke, it multiplies the flow of information — new outputs, automated decisions, reports, generated content — and accelerates the pace of change — new processes, new roles, new tools every week. In other words, it pushes exactly the two forces that destabilize systems.

If, meanwhile, structure and cohesion — the two stabilizing forces — remain where they were last year, the system does not become smarter. It becomes more fragile.

This is the essence of the SICT diagnosis: it does not measure the tool, but the balance.

Four letters, two forces

The four dimensions of SICT are divided into two camps:

Stabilizing forces — these hold the system together:

• S — Structure: the architecture of data, processes, and decisions. Are there clear roles, data flows, decision rights, and governance? Is there a framework that can actually support AI?
• C — Cohesion: shared meaning and trust. Do teams understand the same thing when they look at AI output? Does shared meaning survive the end of the process, or does every department manufacture its own interpretation?

Destabilizing forces — these stretch the system:

• I — Information: AI does not reduce the amount of information; it explodes it. Can the organization separate signal from noise, or does it drown in its own output?
• T — Transformation: the speed of change. Can the team metabolize continuous transformation, or does every new cycle create another crack?

The diagnostic question is always the same: does stabilizing capacity (S + C) keep pace with the load that information and change (I + T) place on the system?

Where this ratio breaks down, the organization does not need more AI. It needs more structure or more cohesion — before the next tool is switched on.

What happens when the ratio breaks down?

The balance can break down in two ways — and both are expensive.

Too much I and T, too little S and C. This is the most common case, and it is the classic “AI hype trap.” The company buys five tools, everyone experiments, output pours out — but nobody knows which number is reliable, who is responsible, or where the whole thing is going. The organization has not become smarter, only louder. This is not progress, but noise accelerated.

Too much S and C, too little I and T. The rarer, but equally real case: the organization protects its own structure and internal consensus so strongly that it watches paralyzed as the market moves past it. Here, governance does not protect — it becomes a cage. AI never moves beyond the “pilot project.”

The point of SICT is precisely that these two mistakes are two endpoints of the same ratio. The goal is not maximum structure, and not maximum innovation. The goal is for your stabilizing capacity to keep pace with the speed you unleash upon yourself.

What does SICT actually measure in an AI project?

Take a typical enterprise team that is now implementing a generative AI solution. SICT translates the vague feeling of “how ready are we?” into four disciplined lines of questioning:

• Structure: Who decides what AI may and may not do? Are the data sources clean, or is the model feeding on garbage? Is there governance, or is everyone using their own shadow tool?
• Information: Do the outputs produced by AI mature into decisions, or do they simply add another layer of noise to the existing chaos? Whose job is it to filter them?
• Cohesion: Are marketing, IT, operations, and leadership talking about the same thing when they say “AI”? Or are they running four different projects under the same name?
• Transformation: Does the organization learn from change, or merely survive it? Do people experience AI as a partner, or as a threat?

The answers do not produce a score. They produce a map: it becomes clear exactly which dimension is the bottleneck — and where the system is under excessive strain.

A diagnosis in practice

Imagine a company of around 300 people — say, a manufacturing or service business. Six months ago, they introduced a trendy generative AI platform. Leadership is proud of it, the press release went out. Then, after six months, someone asks the question: so what has actually changed?

The answer: hardly anything. Let’s run SICT on it.

• Structure: It turns out there is no single owner. IT introduced the tool, but nobody wrote the usage rules. Everyone prompts differently, and the data comes from three different systems. Weak.
• Information: AI pours out material — summaries, outlines, reports — but nobody has time to check what is true and relevant. Overloaded.
• Cohesion: Marketing loves it, operations fears it, middle managers sabotage it because they feel it makes them unnecessary. Four different “AI stories” live side by side. Cracking.
• Transformation: Because the foundation — S and C — is weak, every new feature increases tension instead of creating value. Not metabolized.

The diagnosis is clear: the tool is not the problem. The bottleneck is cohesion and structure — the two stabilizing forces. The company does not need a sixth AI tool. It needs owners, a shared language, and a way to turn middle managers from saboteurs into allies.

This is the real value of SICT: it shows where you should not pour money and energy — and where you should.

How Miklós Róth uses it

Miklós Róth applies SICT not as a mystical formula, but as a diagnostic lens when mapping enterprise AI maturity. This is a deliberate choice: the framework is openly presented as an early-stage systems science proposal, not as a proven law or a universal prediction engine. And that is exactly what gives it credibility.

In practice, this means three things:

1. Mapping the ratio. He does not ask, “Do you have AI?” He asks whether your stabilizing forces — structure and cohesion — can withstand the load that information and change are placing on you. The diagnosis is this ratio itself.
2. Naming the bottleneck. Among the four dimensions, one is almost always the weak link. In one team, structure is missing; in another, cohesion is weak — people do not trust the output. SICT shows where you need to intervene first.
3. Building a shared language. Perhaps the greatest practical value of the framework is that it builds a bridge between leadership, IT, the business side, and communication. When everyone uses the same four dimensions, the “talking past each other” problem begins to disappear — and that alone increases cohesion.

What should you do with the diagnosis?

The diagnosis is not there for its own sake. SICT is useful because the answer to the question “which dimension is weak?” leads to a clear direction for action:

• If Structure is weak → first, rules, owners, and data sources must be put in order, and only then should scaling begin. A new tool will only make the situation worse here.
• If Information is the bottleneck → filters and prioritization are needed, not more output. The question is not “how do we produce more?” but “how do we make faster decisions from what already exists?”
• If Cohesion is cracking → the solution is human, not technological. A shared language, transparent expectations, and the involvement of middle managers — otherwise every implementation will stall from below.
• If Transformation is stuck → the foundation must be strengthened — S and C — before you release more change into the system.

The logic is always the same: strengthen the weakest link, do not keep polishing the strongest one. Most companies do this in reverse — they invest where they are already good.

The point

AI maturity is not about whether you have the best tool. It is about whether your structure and cohesion can keep pace with the information and change you unleash.

SICT does not promise a magic number. It provides honest questions — and in 2026, the year when AI hype starts running out of oxygen, that is exactly what separates a serious partner from the next bullshit generator.

If you want to know where your own team is overstretched across the four dimensions, that is where the conversation begins.

The SICT framework was developed by Miklós Róth (Roth Complexity Lab, Budapest). On the diagnostic application of enterprise AI maturity at Roth AI Consulting.