Build boring things

Every week, a new AI startup launches with a slick demo and a waitlist. A copilot for email. A copilot for meetings. A copilot for your copilot. Most of them will be dead within eighteen months. The ones that survive won't be the ones with the most impressive demos. They'll be the ones that chose to build something boring. Boring is a moat. And in the middle of an AI gold rush, it might be the only moat that matters.

The copilot graveyard

The AI boom has produced an avalanche of products that look incredible in a two-minute screen recording and fall apart the moment a procurement team asks about uptime SLAs. The pattern repeats: a founder builds something flashy, gets traction on social media, raises a seed round, then discovers that Twitter likes don't convert into enterprise contracts. Forbes reported in early 2026 that seed-stage AI startups flashing record revenue numbers were often built on pilot budgets, not real adoption. When those pilots came up for renewal, buyers applied cost-benefit scrutiny they never applied at purchase. A startup reporting $5 million in ARR based on enterprise pilots might find fewer than half of those customers renewing at full price. This is the demo-to-revenue gap. It's the distance between "wow, that's cool" and "we're going to wire you money every month for the next three years." Flashy closes the first sale. Boring closes the renewal.

Why boring works

Look at the companies that print money quietly. Stripe processes payments. Cloudflare routes and protects internet traffic. Plaid connects bank accounts. None of these were exciting at launch. Stripe's original pitch was essentially "we make it easy to accept credit cards online." Cloudflare started as a way to block bad bots and speed up websites. Plaid just moved data between banks and apps. What they share is a set of characteristics that make them almost impossible to displace: Low churn. When your product is infrastructure, customers don't leave because they don't think about leaving. Nobody wakes up and says, "I should switch payment processors today." The product becomes invisible, which is exactly where you want to be. High switching costs. Stripe is woven into codebases. Cloudflare sits between every user and every server. Plaid is connected to millions of bank accounts. Ripping any of them out is a multi-quarter engineering project that no sane CTO would approve unless something was catastrophically broken. Compounding advantages. Every new customer, every new integration, every new data point makes the product stronger. Stripe's fraud detection improves with every transaction it processes. In early 2026, Stripe reported that its payment models boosted fraud detection accuracy by 38 percentage points overnight, a direct result of the data moat it had been quietly building for over a decade. Cloudflare's annual revenue hit $2.17 billion in 2025, growing at nearly 30% year over year, with Q4 revenue surging 34%. That's not hype-driven growth. That's the compound interest of being deeply embedded in how the internet works.

Boring in the AI era

If you're looking for where to build something durable in AI, skip the consumer-facing chatbot and look at the parts of the stack that nobody wants to touch: Data pipelines. Getting messy enterprise data into a shape that AI models can actually use. It's tedious, unglamorous work, and every company with an AI strategy needs it done well. Compliance automation. Regulated industries need to prove that their AI systems follow the rules. Morgan Stanley has already turned to automation for compliance work, and that's just the beginning. The regulatory surface area for AI is only going to expand. Integration middleware. The average enterprise runs hundreds of SaaS tools. Someone has to make them talk to each other. AI makes this both harder (more tools, more data flows) and more valuable (better integrations enable better AI). Monitoring and observability. When AI systems make decisions that affect revenue, someone needs to know when they break. This is plumbing. It's also mission-critical. None of these will get you trending on social media. All of them will get you multi-year contracts with companies that can't afford to churn.

Distribution beats product

There's another reason boring wins: boring things are easier to distribute. When you solve a clear, specific pain point, your sales pitch writes itself. "We make your data pipelines reliable" is a sentence a VP of Engineering can repeat to their CFO. "We're an AI-powered collaborative intelligence platform" is a sentence that makes procurement teams close the tab. The best boring products don't need to educate the market. The problem already exists. The budget already exists. You just need to be the best solution, and then make it very, very hard to leave. This is why distribution matters more than product in most markets. A mediocre product with great distribution will outperform a brilliant product that nobody can figure out how to buy. Boring products have a natural distribution advantage because the buyer already knows they have the problem.

The founder psychology trap

If boring is so obviously better, why doesn't everyone do it? Because founders are builders, and builders want to work on exciting problems. There's a deep psychological pull toward novelty, toward the frontier, toward the thing that makes other engineers say "that's clever." But clever and valuable are different things. The Reddit post that went viral in 2026 about a developer who burned $12,000 on three SaaS ideas (an AI note-taking app, a social media scheduler, a project management tool) while a friend cleared $180,000 running a pressure washing business captures this perfectly. The developer was chasing what sounded cool. The friend was solving a problem people already paid for. The trap is thinking that "boring" means "small" or "uninspiring." It doesn't. AWS started as "rent someone else's servers," possibly the most boring pitch in the history of technology. Amazon launched its first cloud service in 2006, and as Fortune later noted, the real origin story wasn't about spare capacity; it was about rethinking IT infrastructure from the ground up. Today AWS generates more revenue than most countries' GDP. The pitch was boring. The ambition was not.

How to find your boring thing

If you're trying to figure out what to build, here's a simple filter: look for problems where the customer's reaction isn't "wow" but "finally." Look for workflows where people are duct-taping spreadsheets together, manually copying data between systems, or paying consultants to do things that should be automated. Then ask three questions:

If the answer to all three is yes, you've probably found something worth building. It won't make for a great demo reel. But it will make for a great business.

The quiet compounders win

The AI hype cycle will do what every hype cycle does: produce a few massive winners and a graveyard of startups that optimized for attention instead of retention. The winners won't be the ones with the most impressive launches. They'll be the ones who chose the boring problem, solved it well, and made themselves impossible to remove. In a market flooded with noise, the most contrarian thing you can do is build something quiet. Something unglamorous. Something that works so reliably that your customers forget you exist, and keep paying you anyway. Build boring things. Let everyone else fight over the demo.

References