Fewer Developers, More Builders: Why We're Starting an AI-First Project

Three years ago it was a playful weekend experiment: prompting ChatGPT to write a Node.js banner generator, then passing the output to a rival model to critique the code smells — orchestrating two junior engineers in a quiet competition. Today, senior engineering leaders openly say they no longer manually code yet ship more pull requests than ever. That stopped sounding dramatic and started sounding inevitable.

So we committed a real internal product to a deliberately AI-first SDLC: AI generating UI components, scaffolding backend services, assisting testing, drafting documentation — to observe honestly where it accelerates and where it struggles. The industry context matters (a MICE business can experiment faster than regulated insurance, where the process itself is regulated). And a quiet moment during onboarding — explaining Git cherry-picking and feeling it drift from the centre of gravity — crystallised the thesis: when coding becomes conversational, the gatekeeper skill is no longer syntax. We may see fewer developers, but many more builders: designers, analysts, domain experts who understand a problem deeply and can now translate intent into working systems. What AI does not replace is ownership.