Faster summaries. Cheaper code. Smaller teams. All real, all first-order gains, all commoditise within years. The actual reshuffle happens at the second and third order - and almost nobody is measuring those.
Every AI business case starts with the same shape: a list of tasks that get cheaper or faster. That's the first order. It's also the most-talked-about and the most-commoditised.
The second and third order are where industries restructure and value migrates. The first-order winners aren't always the second-order ones. Sometimes they're the losers.
First-order effects are the direct, visible gains from the technology itself. Faster processing. Lower cost per unit. Reduced manual effort. They're easy to measure and easy to attribute. They're also temporary - once everyone has the same tool, the gain commoditises.
Second-order effects are the shifts in power and leverage that the first-order changes enable. A retailer with better data can renegotiate supplier contracts. A warehouse with robots can promise same-day delivery. A firm with AI-coordinated workflows can take on different kinds of work. These changes restructure the industry's pricing and dependence dynamics.
Third-order effects are the structural transformations that follow - which firms exist, which jobs exist, which cities thrive, which industries get redefined entirely. Third-order effects play out over a decade and they're the ones that decide who wins the era. They're also the hardest to forecast, which is why most strategy stops at the first order.
A "faster checkout" tool that restructured global retail.
First-order: 30% faster checkout. Lower cashier errors. Fewer mispriced items. Every retailer that adopted barcodes got this. It looked like a productivity story.
Second-order: retailers now had store-level demand data they'd never had before. Walmart used it to renegotiate supplier terms - demanding tighter delivery windows, standardised packaging, integration into Walmart's logistics. Power shifted from suppliers to retailers. Brands that had dictated shelf placement now had to prove performance per scan.
Third-order: retail itself was restructured. Walmart's barcode-native architecture allowed it to grow into the largest company in the US, restructure where manufacturing happened (toward China), kill Main Street retail in thousands of US towns, and rewire global supply chains around just-in-time delivery. None of that was visible at the cash register in 1975.
Most AI ROI calculations live in the leftmost column. Most strategic decisions about AI should live in the rightmost.
1st order: 30–50% faster code output per developer. Real, measurable, already commoditising. 2nd order: firms that integrate AI deeply into their codebase can ship features competitors can't match - pricing power shifts. 3rd order: the entire engineering job market restructures. Junior roles disappear. Senior IC and product roles reshape around orchestration. Which firms even need an internal eng org becomes an open question.
1st order: 60–70% of tickets handled by AI. Cost per resolution drops. Quick win. 2nd order: customers expect 24/7 instant response - the bar for human-assisted support resets industry-wide. 3rd order: support stops being a cost centre and starts being a direct channel for upsell and intelligence. The firms that figured this out early are now selling "voice of customer" data products at margins their competitors can't see.
What's the 2nd and 3rd order - and which one is your business case built on?
If your business case only works at the 1st order, your competitive advantage has the lifespan of the first-mover window. The durable bets are architected for the 2nd and 3rd.
The 1st-order win pays for the pilot. The 2nd and 3rd order pay for the decade.
The advisory engagement built for the second and third order: clarifying where value migrates in your industry's value stack as AI restructures it, where control points shift, and where the durable advantage emerges.