On artificial intelligence—not as novelty, but as the old failures of leadership and judgement conducted at unprecedented speed and expense.
My guest this week is Jay Fontanini, founder of the AI Executive Accelerator, who spent three decades in property and casualty insurance before turning to building software and teaching executives to think with it rather than perform having thought.
There is a familiar shape to the discussion of artificial intelligence in Australian enterprises. Adoption figures rise. Pilot counts rise. Licence spend rises. Somewhere in the annual report a paragraph
People accountable to a superior whose views they know tend to adopt those views. Nick Hassett on how accountability cultures manufacture silos, why agreement stops at the boardroom door, and why banning AI hides the risk rather than removing it.
An agent, in law and in the older political vocabulary, is someone who exercises judgement on another's behalf and answers for having exercised it. The systems we have agreed to call agents can do the first and not the second, and the evidence on what that costs has started to arrive.
The remedy has no dashboard, vests over years, and mostly benefits your successor—which is precisely why it is avoided. The final essay in the Prior Obligations series sets out what raising the general level of competence actually requires, and why no tool can be bought in its place.
The mandate travels downward and the exemption stays where it was written. The fourth essay in the Prior Obligations series examines organisations that require AI of the crew while the bridge quietly abstains, the consumption quota that enforces it, and the questions a director might ask.
I can audit a model's output on leadership and must take its tax advice on faith. That asymmetry is the whole problem, and it is not psychological. The third essay in the Prior Obligations series argues that AI does not deskill the competent—it unmasks the fake.
Why 'human in the loop' means nothing if the human can no longer tell when the machine is wrong — on judgement, the medieval trivium, and what nine years of flat AI returns are actually measuring.
My guest this week is Jay Fontanini, founder of the AI Executive Accelerator, who spent three decades in property and casualty insurance before turning to building software and teaching executives to think with it rather than perform having thought.
A score creates an owner, a dashboard, a target, and a budget—everything except an improvement in the condition it measures. The second essay in the Prior Obligations series examines what happens when 'psychological debt' becomes a number, and which way its arrows actually run.
Six new names for old wounds. The 'psychological debt' said to accompany AI is neither novel nor peculiar to the technology—it is the perennial business of managing people through change. The first of five essays on what the diagnosis gets right, and what it conveniently overlooks.
There is a familiar shape to the discussion of artificial intelligence in Australian enterprises. Adoption figures rise. Pilot counts rise. Licence spend rises. Somewhere in the annual report a paragraph
An executive search brief gets the diagnosis right: the constraint on value is not technical. Then it advertises a vacancy. The authority to prioritise, to fund, and to stop already sits with someone. Creating a post rather than confronting an incumbent is a governance evasion.
Generative AI has made plausible ideas nearly free, and a fashionable argument says this changes everything. It doesn't. Ideas were never scarce—judgement was, and remains so. On Grub Street at industrial scale, AI slop, and the only mile that was ever hard.
Part IV of a IV part series. The organisations capturing real value from AI are not better at AI; they are better at running organisations. The specific moves boards and executives can make to join them—and why nine years of evidence says most will not.
Part III of a IV part series: thirty-four per cent of organisations claim to be deeply transforming with AI. Eighty-four per cent admit no jobs have been redesigned. These two findings cannot both be true in any operationally meaningful sense. One of them is theatre.
A double issue. April and May kept returning to one move: swapping something legible—a competency grid, a diagnosis, a silver-bullet hire, a perpetual pilot—for the harder work of judgement. The task is always to weigh the particular case, not reach for the proxy that spares the weighing.
Part II of a IV part series: twenty-five per cent of organisations have moved AI experiments into production after nine years of trying. Fifty-four per cent expect to within three to six months—a prediction that has appeared in every major survey for nearly a decade.
