What happens when AI commoditizes?

The book's professional-services argument is direct: 'if AI performs more of the underlying work, but the firm continues to carry the full burden of accountability, the economics and operating model of professional services will come under pressure.' Once AI commoditizes the underlying expertise, the tool provider - having learned from cross-customer use - moves into the solution layer. Commoditization is the gating condition that makes vertical encroachment economically viable.

The book frames the irony of dependence explicitly: 'when the cost of executing knowledge work collapses, and what used to take a team of analysts now takes a single prompt, companies that adopt these tools might gain short-term efficiency but risk losing long-term control.' Commoditization at the task level enables encroachment because, once executed cheaply, the underlying capability is no longer the moat. The tool provider takes the surplus.

When task-level AI commoditizes, expertise that used to define an industry's gates becomes available off the shelf. The book is direct: 'industry boundaries are constantly shifting, making it easier to borrow capabilities from other sectors… A startup from another sector can become your fiercest rival.' MrBeast in burgers, fitness apps cross-selling therapy, insurers pricing health off fitness data - boundaries dissolve because the bottlenecks that held them in place are no longer scarce. Commoditization is the dissolver.

Vertical encroachment is what springs the trap. As the tool provider goes deeper - 'continuously absorbing data, learning from use across a broad set of customers, and integrating those learnings into the core tool' - solution providers reorient their workflows and business models around the engine to extract its performance gains. The book is explicit: the trap is created by 'solution providers marching downward into the tool, hoping to gain its benefits to compete better in the short term.'

Boundary dissolution doesn't restructure value on its own - it creates the conditions under which a new division of labor and a new division of surplus can form. The book's Apple/iPod case captures this: as the recorded-music industry's boundaries dissolved against consumer electronics, Apple rebundled discovery, access, and playback across the boundary, and 'value created in the traditional music industry… was driving value capture in the consumer electronics industry.' Architecture shifts once boundaries cease to hold.

Whoever owns the right to coordinate ends up holding the chokepoint others have to work through. The book traces this through the tool-integration trap directly: as the AI engine improves and solution providers re-orient around it to gain its performance gains, the engine acquires Right to Coordinate, and 'every improvement in the engine delivers an improvement in the solution… but each lift tightens the grip.' Ownership of coordination at workflow scale is what tips solution providers into the trap.

The book draws on Jacobides' point that bottleneck location determines value capture, and applies it to AI: 'when companies misread where to play, anchoring themselves in a system that's fading, they inevitably misread how to win within it.' When industry architecture changes - value chain disaggregates, division of labor shifts, customer journey rebundles - the binding bottleneck migrates. Whatever was the chokepoint in the old architecture (manufacturing access, distribution, expertise) is no longer the chokepoint in the new.

Wrapper position is the third stage of the integration trap, as the book lays it out: first the solution provider builds workflows around the engine; then it reorients its business model to leverage the engine; finally it no longer competes on its own capabilities but on how well it integrates someone else's. At that point the company may still look like a solution provider, but it has become 'a pejorative reference to an interface wrapped around someone else's performance layer.'