The Shape of an Autonomous System
Systems that keep themselves healthy tend to share a shape, and it is older than software. The cybernetician Stafford Beer spent a career on it. His Viable System Model describes any organization that survives as one that senses its environment, decides what needs attention, and acts to correct itself, continuously. W. Edwards Deming drew a simpler version of the same idea, the plan, do, check, act loop, which turned quality control into a cycle rather than a report.
Autonomous business intelligence is that shape applied to a business, run by a system instead of remembered by a person. The loop has a few plain steps. It watches the business as it runs. It detects what changed enough to matter, which is mostly the discipline of ignoring what did not. It investigates, working out why the thing that changed, changed. It recommends the next action. It verifies whether the action actually moved anything. And where it is trusted to, it acts.
The important word is loop. A report ends. You read it and it is over, and the next one starts from scratch. A loop does not end. It runs, and each turn feeds the next, so the system is not answering a question once but keeping a continuous account of the business and its own effect on it.
This is also where the inversion becomes concrete. In the old model, the human drives every turn of the loop by hand: I decide to look, I look, I think, I act. In an autonomous system, the loop runs on its own, and the human steps in at the one turn that needs judgment, the decision. The person is not doing the watching and the sifting and the tracing. They are making the call.
Two cautions keep the idea honest. Autonomous does not mean unattended; a loop that can act still answers to a person. And verifying that something moved is not the same as proving you caused it, which is genuinely hard and worth never overstating. The shape is old and well understood. What is new is that a system can now run it over a whole business without a person turning the crank. That is what the category is, reduced to its motion.