Dr. Jonah Tebaa, a distinguished expert in AI strategy, identifies a critical oversight in how many organizations approach the integration of artificial intelligence: a fundamental misassignment of authority that he terms the "Delegation Gap." Despite significant investments in AI technologies, Dr. Tebaa observes a recurring pattern where decision-making remains slow, senior personnel are burdened with inappropriate tasks, and trust in AI outputs is undermined. His analysis suggests that executives often delegate authority to AI in precisely the opposite way that would yield optimal results.
The core of Dr. Tebaa’s argument is that AI is frequently tasked with ambiguous, high-stakes judgment calls, while humans are retained for pattern recognition — the precise inversion of where each performs best. The machine is handed the decisions that depend on context, relationship history, and factors nobody has written down; the person is left checking outputs a model could verify to 95% accuracy before breakfast.
The gap does not emerge from carelessness, in Dr. Tebaa’s reading. It emerges from three predictable pressures. Speed pressure: automation projects get funded to save time, and the easiest things to automate are the ones with the clearest inputs and outputs — which are also the tasks with the clearest right answers. Teams under deadline automate the measurable and leave the complex untouched. Liability displacement: when a judgment call goes wrong, executives want to be able to say a human reviewed it, so people are kept nominally in the loop on decisions the system in front of them is actually making. The accountability is theatrical. Confidence misattribution: fluent, well-formatted output feels authoritative, and a model that writes a confident recommendation is trusted further than its track record warrants. The polish of the output gets mistaken for the quality of the judgment.
The liability-displacement pressure now has a legal answer. The EU AI Act's Article 14 on human oversight requires a high-risk system to be provided so that the person to whom oversight is assigned is able "to decide, in any particular situation, not to use the high-risk AI system or to otherwise disregard, override or reverse the output". Oversight there means the authority to override, not the presence of a reviewer — the same line Dr. Tebaa draws when he calls the accountability theatrical.
These pressures do not resolve on their own; they require a deliberate framework. Dr. Tebaa maps AI delegation across two variables. Output Variance is how much the correct answer shifts with context, nuance, or relationship history — low-variance tasks have stable right answers, high-variance tasks require situational interpretation. Stakes Accountability is who bears the consequence of a wrong call and how visible it is — low-stakes errors are recoverable and contained, high-stakes errors reach clients, revenue, trust, or legal standing in ways that are hard to reverse. Plotting any task against those two axes is what tells an executive whether it belongs to the machine or to a person.
