Nine years of rising conversation about artificial intelligence, and a measured return that has not moved. The comfortable reading is that the technology was oversold. The uncomfortable one is that we have not yet done the work it requires of us.

For the better part of a decade the surveys have promised imminent transformation, and for the same decade the earnings impact has stayed flat. Jay Fontanini's argument is that this is not evidence of a fraud but the signature of a general-purpose technology mid-adoption — the dynamo was ready long before the factory was rebuilt around it. The complementary investment that matters most, he contends, is not better software. It is human capital, and at the centre of that, judgement.

Which raises the harder question the conversation works: if the models are now a near-substitute for the quick reasoning and rapid recall that peak early in a career, then the entire human premium falls on the accumulated, seasoned kind of intelligence that cannot be hurried into existence. Jay teaches executives through the structure of the medieval trivium on the principle that no stage can be skipped — and reports that they do not give up at the difficult foundations, but at the middle, where thinking with the tool actually begins.

Takeaways

• The flat productivity number is a phase, not a verdict — but a theory that explains any delay as unfinished co-invention is only honest if the complementary work is actually being done.
• Artificial intelligence substitutes for fluid intelligence, not crystallised. It is cheapest exactly where we were quickest, and no help at all where we were wise.
• Most AI-attributed layoffs are ordinary corporate bloat wearing a new excuse — and they sit in direct contradiction with the efficiency dip that genuine adoption requires.
• Executives do not abandon the learning at the grammar stage. They abandon at the dialectic, and leap to holding forth about a tool they have not learned to think with.
• Judgement is earned rather than learned: cumulative, progressive, and not compressible into a prompt library.
• Friction belongs in the workflow by design. A system may draft anything; it should be permitted to send nothing.
• The engineers built a language model, which puts the humanities on the ground floor of the most technical instrument in the building.

Chapters

Guest Links & References

• Writing and notes: jayfontanini.com
• LinkedIn: linkedin.com/in/jay-fontanini/
• Prediction Machines: The Simple Economics of Artificial Intelligence — Ajay Agrawal, Joshua Gans and Avi Goldfarb (2018). Jay's nomination for the one book to read before touching a prompt.
• From Strength to Strength: Finding Success, Happiness and Deep Purpose in the Second Half of Life — Arthur C. Brooks (2022), the source of the fluid and crystallised intelligence framing discussed in the episode.
• The Diversity Bonus: How Great Teams Pay Off in the Knowledge Economy — Scott E. Page (2017), on complementary skill in problem-solving teams.
• Superintelligence: Paths, Dangers, Strategies — Nick Bostrom (2014)

About the Show

On the Subject of Leadership is a long-form interview series, with a written companion, on governance, organisational culture, and how decisions are actually made. Guests are challenged rather than affirmed, and the argument continues in writing after the recording stops. For people who are sceptical of easy answers.

Hosted by Dr Robert N. Winter.

Subscribe / Follow

Newsletter / Website: robert.winter.ink

LinkedIn: dr-robert-winter

X: @DrRobertWinter

Instagram: DrRobertWinter

Mastodon: social.winter.ink/@robert

YouTube: @OnTheSubjectOfLeadership

Credits

Recorded remotely via Riverside

Music: The Hidden Thread by Roberto Prado / Artlist