RESHUFFLE An interactive companion to the book
Explainer · Chapter 8 ▸ Strategy cluster

The AI tools you depend on will eventually compete with you.

Every integration delivers performance gains and moves pricing power one notch upstream. By the time you notice, you're a wrapper around someone else's engine.

The seductive framing is “we'll just adopt best-in-class AI tools.” That gets you a productivity bump and a steadily accumulating dependency.

Five years in, the tool that powers your customer promise belongs to someone who can - and will - eventually serve your customers directly.

▍ The mechanism

Tools vs engines - and the three forces that always tilt upstream

A tool performs an isolated function - you can swap it. An engine determines system performance - the rest of the system is built around its capabilities. Tools get bolted on. Engines require integration. The choice between treating an AI capability as a tool or as an engine is the strategic decision.

Three structural forces always tilt the value chain toward whoever provides the engine. Learning - the provider learns from every customer's operations; you learn only from your own. Scope expansion - tools broaden horizontally into adjacent capabilities and vertically into your business. Clockspeed - the AI layer ships new capability every week; your industry ships new capability every quarter at best.

The end state is the wrapper. The engine becomes central to your performance. You reorient operations around it. Internal differentiation atrophies. Eventually customers aren't buying your expertise - they're buying an experience powered by the same engine that powers your competitors. Performance-based lock-in is harder to escape than contractual lock-in because leaving means falling behind the industry.

▍ Historical analogue

Renault vs Ferrari, 2005 Bahrain GP

When buying the best engine stopped being enough.

Ferrari had dominated Formula 1 for half a decade. Bigger engine. Bigger budget. Schumacher untouchable. At the 2005 Bahrain Grand Prix the trackside thermometer pushed past 40°C and Schumacher's tires began to blister. The car - built as an assembly of best-in-class components - couldn't adapt to conditions its engine wasn't tuned for. Mid-race, the engine failed.

Alonso's Renault won. Not because Alonso outdrove Schumacher, but because Renault had built the entire car as an integrated system around their own engine - chassis, gearbox, aerodynamics, tire degradation, fuel curve, all tuned to each other. They owned the engine; the rest of the system was designed for it.

From that moment, winning F1 teams stopped chasing the best components and started owning the engine. Mercedes, Red Bull, Ferrari all became their own engine suppliers. The teams that kept buying components are still on the grid. They're not winning championships.

This is the same fork sitting in front of every firm adopting AI.

▍ Tool vs engine

Two ways to bring AI into your business

Frame A
AI as a tool
Where
Bolted on, alongside existing workflows
Swap?
Yes - change vendors anytime
Effect
Productivity bump on tasks
Compete on
The same terms as before, slightly faster
Pricing power
Stays with the tool provider
Frame B
AI as the engine
Where
Centre of the business - everything else is built around it
Swap?
No - leaving means re-architecting
Effect
The basis of competition shifts
Compete on
New terms you've defined
Pricing power
Belongs to whoever owns the engine

The trap is that adopting AI looks the same in both frames at year one. The difference shows up at year five, and by then it's expensive to reverse.

▍ Two firms that learned the hard way

The wrapper end-state, twice

01

Uber and Google Maps

Uber started as a happy Google Maps customer - what could be better than free routing intelligence? By 2016 it was clear: every pickup error, every miscalculated ETA was Uber's customer experience. Maps was the engine; Uber was the wrapper.

Uber spent years and millions trying to escape - tried to acquire Nokia's mapping unit, partnered with TomTom and HERE, eventually built proprietary mapping. By their 2019 IPO they were still paying Google significant sums. Once you're in the trap, escape costs years.

02

Netflix and the studios

Netflix as a streaming distributor was hostage to studio licensing. The better Netflix got at acquiring and retaining users, the more leverage the studios had at renewal. There was no path to expanding margins as a wrapper around someone else's catalogue.

The only escape was creating their own content. They became their own engine. The bet cost billions and worked. The firms that didn't make that bet - Quibi, the early streaming aggregators - got absorbed or died. The pattern repeats with AI.

▍ Apply it

Where would your AI bet land?

For each of the six capabilities, decide: do you own it, buy it, or skip it? Your distribution reveals where you sit on the engine-ownership map.

Six capabilities your firm could invest in. Pick a stance for each.

  1. 01
    Foundation modelsRaw LLM / vision / voice capability - the underlying intelligence.
  2. 02
    Domain data & trainingProprietary signal from your industry - what nobody else has.
  3. 03
    Workflow integrationHow AI weaves into your day-to-day operations.
  4. 04
    Customer experience layerThe front-end your customers see and touch.
  5. 05
    Trust, compliance & accountabilityLiability, regulation, the thing customers sue you over.
  6. 06
    Distribution & customer relationshipWho customers think they're buying from. Where the order originates.
▍ Where you land

The Laggard

You're not really playing the AI game. The decision to not adopt is itself a strategic choice - and increasingly an expensive one.

0Own
0Buy
0Skip
6Undecided