The middle of the software industry is in trouble

The software industry is reorganising into three distinct layers, and the companies sitting in the middle are in serious trouble.

Think about what Stripe does. It does not have an opinion about your business. It provides a payment rail. You plug in and build whatever you want on top. Same with AWS, OpenAI, Anthropic. They provide raw capability such as cloud infrastructure, AI engines and building blocks. The more diverse the things people build on top of them, the more valuable they become.

Now think about what Salesforce, SAP, Workday or ServiceNow do. They have very strong opinions about your problem. Salesforce says selling looks like this. SAP says your operations work like this. Workday says HR looks like this. For four decades that was their strength because most people could not afford to disagree. Building your own version was too expensive, too slow, too complex. So you accepted the template and adapted your business to fit it.

That trade-off made sense when the cost of custom was prohibitive. It makes a lot less sense when a small team can build a tailored alternative in weeks for almost nothing. The capability jumps of the past year are what tipped this from interesting trend to structural shift happening now.

Three layers are emerging.

The bottom layer is infrastructure. This consolidates. Fewer companies, larger, more dominant. AWS, Stripe, OpenAI, Anthropic. They win because they are unopinionated. They provide building blocks, not blueprints.

The top layer is niche solutions. This explodes. Thousands of specific tools built by people who understand the problem from the inside. A sales pipeline built for how SaaS teams actually sell. A compliance tracker designed for a specific industry. A client onboarding workflow that matches how your firm actually works. Each one purpose-built, not a generalised version of someone else's idea of your problem.

The middle layer is opinionated platforms. Salesforce, SAP, Workday, ServiceNow, Shopify, HubSpot, Atlassian, Xero. This is where the disruption is most violent. These companies are simultaneously too opinionated to be infrastructure and too generic to be the niche solution. They are selling pre-built structures in a world where everyone can now build their own.

Their power was never really the features. It was being the system of record, the place your data lived, the thing everything else connected to. That made them hard to leave. As AI agents become the primary interface for work, the system of record becomes a layer underneath rather than the thing you interact with. It becomes interchangeable. The lock-in that protected these platforms for decades starts to dissolve.

The middle was always the "best practice" layer. It convinced you your problem was the same as everyone else's. It worked when the economics demanded it. Those economics just changed permanently.

If you are a leader evaluating your technology stack right now, the question is not which platform to back. It is which layer each part of your stack actually sits in, and whether the middle layer you are paying for is about to be replaced by something built specifically for you.

How does your organisation actually score?

Get a free AI Disruption Snapshot in five minutes — no credit card.