The Swiss Watch Moment for Software: How AI Is Commoditising Code (And What Survives)

What the quartz crisis did to Swiss watchmaking, AI is doing to manual coding. The craft isn't dead — but the market for it is splitting in ways the industry hasn't fully processed yet.

· 5 min read
ai-engineering software-industry engineering-culture ai-adoption future-of-work

Key takeaways

  • The quartz crisis didn't kill watchmaking — it killed the middle. Premium Swiss mechanical watches survived. Mass commodity watches moved to Asia. There was no stable middle ground.
  • AI-assisted coding is doing the same thing to software: commoditising the production layer, while the architectural and judgement layer remains genuinely scarce.
  • IT services built on labour arbitrage — billing for hours of code production — are in the position of mid-tier Swiss watchmakers in 1975. The math stops working.
  • The developers and firms that survive this transition will be the ones who stop selling code production and start selling outcomes, judgment, and leverage.

I’ve been thinking about the Swiss watch industry a lot lately.

Not because I collect watches — though I do find the mechanical ones disproportionately beautiful for something that a cheap quartz movement outperforms on every objective measure. But because what happened to that industry between 1969 and 1988 is the clearest historical analogue I’ve found for what’s happening to software right now.

The quartz crisis

In 1969, Seiko unveiled the Astron at the Tokyo Olympics — the world’s first commercially available quartz watch. It was accurate to within five seconds per month. The best Swiss mechanical watches of the era were accurate to within a minute. The quartz movement required no skilled craftsman to assemble, no annual servicing, and cost a fraction of what a Swiss movement cost to produce.

The Swiss watch industry had roughly 1,600 manufacturers at the time and employed 90,000 people. By 1988, those numbers were 600 manufacturers and 28,000 employees. The middle had collapsed.

What’s interesting is what survived and what didn’t. Mass-market watches moved to Japan and Hong Kong almost entirely — Seiko, Casio, Citizen. Swiss brands that tried to compete on volume and price were destroyed. The ones that survived did one of two things: they moved aggressively downmarket with Swatch (cheap, fashionable, Swiss-made but designed for disposability) or they moved aggressively upmarket into mechanical watchmaking as craft, heritage, and status object — Rolex, Patek Philippe, IWC.

The Swiss watch industry that emerged from the crisis is smaller, higher-margin, and more differentiated. It doesn’t compete on telling time. Nobody buys a CHF 20,000 mechanical watch because it’s more accurate than a CHF 50 Casio. They buy it for what it represents: craft, longevity, the fact that it was assembled by a human being with skills accumulated over a lifetime.

The middle — good enough quality at mid-range prices, competing primarily on craftsmanship — largely ceased to exist.

The parallel

In 2020, writing code was a craft. You needed years of practice to do it reliably. Senior engineers commanded significant premiums because their accumulated judgment reduced risk. The cost of building software was high because the supply of people who could do it was constrained.

AI coding tools didn’t eliminate that, but they did something structurally significant: they decoupled code production from engineering judgment.

A reasonably capable developer with Claude Code or Cursor can now produce code at a rate that would have been impossible five years ago. The mechanical work — translating a specification into working syntax, implementing a known pattern, writing tests for a defined interface, refactoring to a different convention — is increasingly automated. Not perfectly, not without oversight, but well enough to reshape the economics.

The cost of code production is moving toward zero. Not today, not completely, but directionally and at speed.

What this does to the industry is what the quartz movement did to watchmaking: it collapses the middle.

What collapses first

The most exposed segment is IT services built on labour arbitrage.

The model worked like this: you hire engineers in lower-cost geographies, bill them to clients at a margin, and the value proposition is approximately “we produce code cheaper than you can in-house.” The metric is hours. The incentive is to keep people busy, because people-hours are the product.

AI tools don’t fit this model at all. They reduce the number of hours required to produce a given amount of code. The client benefits. The firm that bills by the hour does not.

The analogy is a mid-tier Swiss watchmaker in 1975 trying to respond to Seiko by optimising their assembly process. Optimisation isn’t the answer. The business model is what’s wrong.

Consulting firms built on bespoke software delivery face a version of the same problem. If your differentiation is “we build things carefully and it shows in the quality” — and AI tools allow any team to produce careful, well-structured code much faster — what exactly is the premium for?

What survives

The Patek Philippe end of the market is architectural judgment.

Building software at scale — choosing the right data model for a product that needs to grow ten times, designing an API contract that a distributed team can work against without stepping on each other, knowing when a microservices architecture will cause more problems than it solves and when it won’t — this is not what AI tools are replacing. They can assist it, but they can’t substitute for the judgment that comes from having made consequential architectural decisions and lived with the results.

The Swatch end of the market is speed and accessibility.

AI-assisted development is genuinely democratising software creation. Non-technical founders are shipping MVPs. Small businesses are building their own internal tools. Things that would have required a six-month engagement with a development firm can be put together in days. This is mostly good. It creates a market that didn’t exist before.

The middle — “we write code professionally and it’s good” — is where the pressure concentrates.

What this means for developers

The engineers I’ve seen adapt fastest to this shift are the ones who updated their mental model of what they’re selling.

They’re not selling code production. They’re selling the ability to ask better questions, catch the failure modes that AI confidently introduces, hold architectural context across a codebase that’s now growing faster than it ever has, and know when to override the model.

That’s a legitimate and scarce skill. It just requires letting go of the identity that was built around writing code by hand.

The Swiss watchmakers who transitioned into the luxury segment didn’t stop valuing craft — they found a market where craft was the point again, not a production method that happened to produce reliable outputs.


If your organisation is working through what AI adoption means for your engineering team — commercially, structurally, and culturally — let’s talk.

Free Assessment

Is your team ready to act on this?

Find out where your engineering organisation stands across 7 dimensions — AI adoption, testing, culture, governance, and more. Takes 8 minutes.

Frequently asked questions

Will AI replace software developers?
The more precise question is: which parts of software development will AI commoditise, and which parts require human judgment? Code production — writing functions, implementing specifications, refactoring — is being commoditised. Architecture, strategy, product thinking, and the ability to ask the right questions are not. The role evolves; it doesn't disappear. The Swiss watchmakers who asked 'will quartz replace mechanical watches?' missed the better question: 'what will mechanical watches be for?'
How should software consulting firms respond to AI commoditising code production?
The firms that adapt fastest are the ones repositioning from selling effort to selling outcomes. Instead of billing for hours of development, they bill for delivered capabilities, transformation results, or embedded expertise. This requires a different commercial model and different skills — but it mirrors exactly what the premium Swiss watch industry did: stop competing on volume and compete on craft, trust, and judgment.
How long will the transition take?
The quartz crisis played out over roughly fifteen years (1969–1988). My read is that the AI transition in software is moving faster — the tooling is improving exponentially, the cost of adoption is low, and the competitive pressure is global. Teams that are still debating whether AI matters in 2026 are already three years behind. The window to adapt is shorter than most people think.
Portrait of Rajesh Prabhu

Written by

Rajesh Prabhu

Fractional CTO & Founder

Rajesh Prabhu is the founder of Seven Technologies and 124Tech. He specialises in AI-first engineering, Harness Engineering methodology, and helping teams operate at a fundamentally higher level of leverage with AI tooling.