Uploaded July 2026 | Updated November 2026, 3 hours ago
In 1979, an internal IBM slide read: "A computer can never be held accountable. Therefore a computer must never make a management decision." Forty-seven years later, the infrastructure industry is preparing to vest that very authority in AI agents, deployed atop environments whose own operators cannot reliably describe what is in them. IBM's observation never lost its relevance, yet rarely informs the decisions now clearing AI for that role.
Decades of extraordinary acceleration in execution capability produced no corresponding advance in the ability to comprehend what infrastructure represents, where accountability resides, or what constitutes a known-good state. That asymmetry has become structural: environments evolve faster than any organisation can comprehend, govern, or secure. Autonomous AI is about to inherit that condition wholesale.
At the centre of this talk sits a falsifiable proposition: the binding constraint on safe infrastructure automation is systemic legibility. Closing that gap requires an operating model, a holistic governing abstraction that renders infrastructure comprehensible at enterprise scale, encodes ownership and intent, and continuously reconciles what an environment is supposed to look like against what it actually does. This argument is drawn from enterprise deployments where the framework has been subjected to genuine operational pressure, and where the outcomes, including the instructive failures, are examined without embellishment.
A far more consequential implication concerns AI itself. Autonomous agents reasoning over fragmented telemetry, inferred topology, and undocumented intent will produce hallucination at infrastructure scale. This is, at its foundation, a crisis of data authority rather than a frontier problem in machine learning. The talk will examine where the
boundary ought to fall between AI that advises and AI that is permitted to act autonomously, and why the organisations that miscalculate that boundary first will define the terms of failure for an entire industry.
#AutoCon5 #artificialintelligence #ai #infrastructure
In 1979, an internal IBM slide read: "A computer can never be held accountable. Therefore a computer must never make a management decision." Forty-seven years later, the infrastructure industry is preparing to vest that very authority in AI agents, deployed atop environments whose own operators cannot reliably describe what is in them. IBM's observation never lost its relevance, yet rarely informs the decisions now clearing AI for that role.
Decades of extraordinary acceleration in execution capability produced no corresponding advance in the ability to comprehend what infrastructure represents, where accountability resides, or what constitutes a known-good state. That asymmetry has become structural: environments evolve faster than any organisation can comprehend, govern, or secure. Autonomous AI is about to inherit that condition wholesale.
At the centre of this talk sits a falsifiable proposition: the binding constraint on safe infrastructure automation is systemic legibility. Closing that gap requires an operating model, a holistic governing abstraction that renders infrastructure comprehensible at enterprise scale, encodes ownership and intent, and continuously reconciles what an environment is supposed to look like against what it actually does. This argument is drawn from enterprise deployments where the framework has been subjected to genuine operational pressure, and where the outcomes, including the instructive failures, are examined without embellishment.
A far more consequential implication concerns AI itself. Autonomous agents reasoning over fragmented telemetry, inferred topology, and undocumented intent will produce hallucination at infrastructure scale. This is, at its foundation, a crisis of data authority rather than a frontier problem in machine learning. The talk will examine where the
boundary ought to fall between AI that advises and AI that is permitted to act autonomously, and why the organisations that miscalculate that boundary first will define the terms of failure for an entire industry.
#AutoCon5 #artificialintelligence #ai #infrastructure










