Picture this: you're in a meeting, and someone mentions needing a feature flag system. Before anyone can suggest LaunchDarkly or Split, Bob from engineering pipes up: “I can just ask ChatGPT to build us something simple this afternoon.” Everyone nods enthusiastically. After all, why pay for expensive third-party tools when AI can whip up exactly what you need in a few hours?
Boom—welcome to "Not Invented Here" syndrome on steroids.
“Not Invented Here” (NIH) syndrome is that voice in your head that says, “we can build that ourselves,” every time someone suggests using an external solution. It's the organizational equivalent of refusing to ask for directions because you're pretty sure you can figure out the route—even though your phone has GPS and traffic data that would get you there faster.
For decades, this syndrome has led engineering teams down expensive rabbit holes: rebuilding open-source libraries, creating internal project management tools instead of using Jira or Linear, building custom authentication systems instead of integrating with Auth0. The drivers are always the same: technical pride, desire for control, and the belief that your requirements are uniquely special.
But here’s the twist: AI just made this problem way worse.
AI coding assistants have fundamentally changed the build-vs-buy conversation. The “just ask ChatGPT to build it” mentality is everywhere. Junior devs scaffold apps in hours. Senior engineers prototype features they would’ve laughed off five years ago.
The barrier to entry for custom development has seemingly vanished. What once required deep domain knowledge now just needs a well-phrased prompt.
AI excels at creating functional prototypes that nail the obvious requirements. Ask it to build an internal survey tool? Boom—basic form builder with submission handling. Need a file uploader for support requests? Done—drag-and-drop interface with backend storage and notifications.
You start thinking you’re basically a software wizard.
But here’s the reality check: AI solves the easy part. Building a basic version is like assembling IKEA furniture—it looks good when you first put it together, but it starts wobbling after a couple of moves and doesn’t age well.
Maintenance Debt. Your AI-generated auth system will need security patches. Your custom CMS will break with the next React upgrade. Your feature flag tool will need scaling when traffic spikes.
Every custom solution becomes a permanent resident in your technical debt apartment complex—and the rent never goes down.
Feature Creep. That “simple” feature toggle system? Soon you’ll need user targeting, percentage rollouts, environment configs, and audit logging.
Before long, your weekend project is a distributed system requiring full-time support.
Opportunity Cost. Every hour building commodity features is an hour not spent building differentiators.
The Bus Factor Problem. When Sarah, who built your AI-assisted tool, leaves, what happens? AI helps with code reading, but not with context, edge cases, or team tribal knowledge.
Testing Is Still Hard. AI can write unit tests. But understanding business logic and edge cases? That still needs human intuition and solid QA practices.
Integration Hell. Your homegrown feature flags have to integrate with deployments, monitoring, user management, and more. AI might generate the pieces, but stitching them into a cohesive architecture is still hard.
🔥 Pro Tip: If it looks easy to build, it’ll be harder to maintain.
Sometimes, custom makes sense. The trick is knowing when.
Ask: Is this our special sauce, or just table stakes?
Netflix’s recommendation engine? Special sauce. Their login system? Table stakes.
Uber’s pricing algorithm? Special sauce. Their credit card form? Table stakes.
Spend engineering time on what makes you money. Buy everything else.
Custom development feels easier than ever—but the economics haven’t changed.
The best companies focus their engineering firepower on core differentiators and rely on trusted tools for everything else.
Next time you're tempted to build something custom, pause. Ask:
Use your AI-enhanced powers wisely. Build what matters. Buy the rest.
Knowing where to invest can be the difference between success and failure.
Want to read more about table stakes features? Check out this related post: The Hidden Complexity of Table Stakes Features.
This post is also part of a recurring theme we highlighted a few years ago, in the post Work Is Work. Just because something seems easy (or even is easy) doesn't mean that it doesn't take time, effort, money, and attention.
