The Goal vs The Hype

You may have heard of the Theory of Constraints (TOC), a way of thinking developed by Eliyahu Goldratt in the early 80s. It was born in manufacturing lines, but applies to almost any area.

TOC treats a system like a chain. Each step is a link, and throughput is limited by the weakest link: the step with the least capacity. Polishing the other links won’t increase overall output if you don’t fix the bad one. TOC is about finding that constraint, making the most of it, and then upgrading it.

I’m simplifying, of course. Goldratt wrapped this into The Goal, which manages to somehow be both a business book and a novel.

Anyway, every time I see AI-driven or AI-assisted development being celebrated in organisations for its “productivity”, it reminds me of the TOC. The constraint is rarely number of lines of code produced per hour. And when AI does accelerate coding, the bottleneck usually relocates elsewhere: code review, testing, refactoring, security, SRE, governance, or even upstream to requirements clarity.

This shows up most shamelessly when a vendor or agency sells itself as “AI boosted” (an actual term in a proposal I read last week). What you need to figure out in that case is what happens after the code is generated.

How are review, security, testing, and long-term maintenance handled? Faster code production is not the same as faster delivery of reliable software. If your delivery timeline shrank, something got moved, deferred, or even dropped.

Colin