17/03/2026
You are not doing automation with AI You are not doing automation with AI; you are doing delegation. This is fine to a certain point, but it is crucial to understand the distinction. Relying on AI does not automate the work; it shifts the reliance to get the job done from a human to an AI agent. This is an agentic workflow that could be done by a human or an AI, but it is not true automation.
True automation is a Finite State Machine (FSM). It does the work reliably, deterministically, and repeatably. It is not subject to hallucination, probabilistic variance, or human mood.
It is counter-intuitive, but real automation is simple. Think of it like a traffic light: if the timer expires, the light changes. There is no guessing. Think of a plumbing system: Water comes in and gets funneled to the network. No flooding, no manual work. It is reliable, simple, cheap, and hard-coded to do one thing well.
The Architecture of Intelligence
We can look at the architecture of any intelligent system (like the human brain) to understand this. There are roughly two layers:
The Conscious Layer: This involves research, thinking, decision-making, and exploration.
The Subconscious Layer: This involves routines and automatic responses (reflexes). It processes signals from the environment without thinking. Driving a car or reacting to an imminent threat falls here. These routines are triggered by an internal or external signal and start processing automatically.
The Takeaway: Everywhere a decision-making or thinking process is involved, it is not automation. It is "AI in the loop" or "Human in the loop."
When you shift reliance from a human to an AI, you are delegating, not automating, because you are still relying on a thinking agent. True automation is building Finite State Machines that do the job without needing human or AI intervention.
Effectiveness vs. Capability
Just because a human can do a task does not mean it is the most effective way to do it. The same applies to AI. If AI can do it, it does not mean it is the most effective way to do it.
To build true automation, we must adhere to strict criteria. We must not rely on probabilistic outputs, hallucinations (AI), or human moods. The criteria for true automation are:
Reliability: It works every time.
Repeatability: It produces the same result every time.
Cost Efficiency: It costs less to run than human labor.
Energy Efficiency: It consumes minimal resources.
Testability: It can be rigorously tested and verified.
The Future of Software
With that said, I believe there will be more software than ever before. AI and humans will work together to build custom software solutions for most of our daily usage, fixing problems as they arise.
However, we will not see a surge in generic, VC-backed SaaS products. Instead, we will see custom software built for people and organizations willing to fix and automate their workflows. We will use AI and humans for research, orchestration, and discovery to build these automations, but the ex*****on itself must be deterministic.
