Jean-Michel Lemieux, the former CTO of Shopify, has joined Spellbook, an AI-powered contracts platform, in a role he describes as executive individual contributor. He was offered the CTO title. He turned it down.
The throughput thesis
Lemieux's argument is that the intellectual property of a company is no longer its code. It is how the company is organised and operates. The job is not building the product. It is rebuilding the organisation so the product can move faster.
He uses the analogy of electrification. When factories first got electricity, they replaced steam engines with electric motors and changed nothing else. The real productivity gains came later, when someone realised you could rearrange the entire factory floor because you were no longer tethered to a central power source.
Most companies are in phase one of AI adoption. They have bolted AI onto existing processes. The 10x improvement comes from rearranging everything around what AI makes possible.
Why contracts
Spellbook has 4,500 companies across 80 countries and more than 10 million contracts on its platform. Lemieux sees contracts as infrastructure, the plumbing that determines how fast the economy can move.
The data advantage matters. Ten million contracts give Spellbook a training and evaluation dataset that a new entrant cannot easily replicate. When a new model drops, the company can test it against real-world complexity rather than synthetic benchmarks.
The Shopify question
Lemieux was asked whether AI could disintermediate Shopify itself. His answer was measured but clear: Shopify's obsession with understanding what entrepreneurs need is its moat. Model providers do not care about those needs in the same way.
Related reading
- Shopify beats every estimate Wall Street throws at it, and the stock still drops 13%
- Space data centres are the new frontier, but the launch crunch is real
- Futures Lab students build AI prototypes for learning Japanese and sign-language practice
The implication is broader. Companies that survive AI disruption will not be the ones with the best models. They will be the ones that understand their customers deeply enough that no model provider can replace the relationship.
That is a less exciting thesis than artificial general intelligence. It is probably more accurate.