Your AI transformation will hit month three and fall apart. Technology is good and enthusiasm is at its peak. But you are generating far more complexity than your governance can absorb, and there is a seventy-year-old scientific law that says this was always going to happen.
Picture the week in almost any enterprise:
- Three department heads demo three different AI tools to the same CEO — none of them aware of what the others are doing.
- The IT data privacy document was signed yesterday, and it already describes a tool version deprecated three weeks ago.
- The board asks for an AI ROI number and nobody can produce one, because nobody agreed on what counted as AI spend when the pilots started.
- Your steering committee meets once a month. By the time it sits, four decisions that needed its sign-off already happened without it.
Most will call this a coordination problem and schedule more meetings. The meetings change nothing, because coordination is not what broke.
Ashby and the "Law of Requisite Variety" (1956)
Ross Ashby, a British cybernetician, wrote the mathematical reason in 1956. His Law of Requisite Variety states:
A controller can only govern a system if it holds as much variety as the system itself.
"Only variety can absorb variety."
Variety simply means the number of states a system can be in. A light switch has two. Your organization running fifteen or fifty AI initiatives across five business units has a number no monthly committee can count, let alone control.
Stafford Beer and Management Cybernetics
Stafford Beer took Ashby's law into the corporate world and built management cybernetics around it in Brain of the Firm (1972) and The Heart of Enterprise (1979).
Beer's point is the exact reality executives are living today:
Every channel that carries control in your company — every committee, every report, every governance forum — must itself hold enough variety to match what flows through it. When the channel holds less variety than the activity, control does not slowly weaken. It breaks, and the activity keeps running without it.
A monthly steering committee is a low-variety channel. Your AI activity is high-variety, decentralized, and accelerating. Beer would have predicted the collapse before your first pilot went live.
A Design Problem You Fix With Structure
This is why the executive running the AI transformation feels constantly underwater while doing everything right. They are not disorganized. They are a regulator whose variety got hopelessly outrun by the system they own.
The organization kept adding AI variety (a pilot here, an LLM vendor there, a new board demand every week) and never added matching variety to the governance system meant to control it.
That is not a motivation problem you fix with effort. It is a design problem you fix with structure.
The Gap Called "Month Three"
None of this slows you down. It means your governance has to grow as fast as your AI activity. That almost never happens, because a growing roadmap is visible and rewarded, while growing governance is invisible and costs money.
So the roadmap runs ahead, the governance falls behind, and the gap between them is month three.
Every initiative you launch before closing that gap does not add capability. It just extends the part of your company that nobody can govern anymore.
Ps: The irony is that AI is very good at building the very thing that could govern it. Almost nobody is using it for that.
- Ashby, W.R. (1956), An Introduction to Cybernetics, Chapman & Hall (Law of Requisite Variety).
- Beer, S. (1972), Brain of the Firm, Allen Lane The Penguin Press (The Viable System Model).
- Beer, S. (1979), The Heart of Enterprise, John Wiley & Sons.
Replace slow monthly committees with an autonomous War Room.
brainTerms.ai matches the speed of AI innovation with continuous multi-agent monitoring, cross-functional risk reconciliation, and single-pane roadmap clarity.