Build vs. Buy: Why Customers Building Their Own Software Is a Bigger Threat Than AI Startups

In short: AI is making it much cheaper for companies to build software themselves. Many SaaS CEOs now see that as a bigger threat than AI startups. Experience shows, however, that customers mainly replace simple, interchangeable features. Systems that bring together authoritative data, industry expertise, operations and liability are rarely built in-house. Vendors should therefore figure out which part of their product can be copied and make a fresh case for price and value right there.

How big is the fear of customers building their own?

85%

In a survey by pricing vendor Cruxy of 300 B2B SaaS CEOs in the UK and the US, 85 percent see AI as a direct threat to their business model.

Asked about the bigger danger, 54 percent name tools that customers build themselves, often called “vibe coding,” and 45 percent name AI-native startups (Cruxy survey, April 2026). The figures come from a vendor with its own stake in pricing and can’t be transferred to the German-speaking market. The direction is still striking: vendors increasingly see the danger in their own customers rather than in new competitors.

Why is building in-house suddenly attractive?

Three developments are coming together. AI tools sharply lower the effort needed to develop applications, even for teams without a large engineering department. Many standard products offer more features than a customer uses and charge for them anyway. And price increases at renewal run into procurement departments that now seriously evaluate alternatives. If you pay for a hundred-user license but only need three features, you do the math differently than you did five years ago.

What does the Klarna example show?

In 2024, Klarna announced that it would replace Salesforce and Workday with its own AI solutions. According to OMR Reviews, the reality looked different. Workday was largely replaced by another vendor, and the CRM functions were spread across smaller tools and an in-house knowledge graph. CEO Sebastian Siemiatkowski later admitted: “We did not replace SaaS with an LLM.” (OMR Reviews)

The lesson isn’t that building in-house fails. It’s that building in-house mainly changes the makeup of the software stack. Large suites lose features to specialized tools and internal solutions, and the data moves into in-house structures. For vendors, that’s a slow bleed, not a sudden blow.

What do customers build themselves, and what don’t they?

More likely built in-houseMore likely bought
Simple interfaces and forms on top of existing dataSystems of record with audit trail and permissions
Reports and dashboards for individual teamsSolutions with deep industry expertise and regulatory requirements
Automation of individual, clearly defined workflowsServices with operational responsibility, availability and liability
Features that make a product expensive but hardly anyone usesData and patterns that only emerge across many customers

Everything in the right-hand column gains its value over years, from data, expertise and accountability in practice. That’s exactly what an internal team can’t build on the side. Those who build in-house also often overlook the follow-on costs. They include maintenance and security, adapting to new requirements and dependence on the few people who understand the system.

How should software vendors respond?

Be honest about what can be copied.Which features could a customer build with AI tools within a few months? That’s where prices will be hard to defend going forward.

Shift the value to data, expertise and operations.Price and sales pitch should rest on what is hard to copy, not on features.

Be open to customers building their own.Good APIs and open standards turn your system into a platform customers build on, instead of replacing it.

Review your packaging.Customers who pay a lot and use little are the first to build alternatives. Leaner entry-level packages protect the base.

Frequently asked questions

Are companies replacing SaaS with homegrown AI solutions?

In part. Simple features, interfaces and reports are increasingly built in-house. Systems with authoritative data, industry expertise and operational responsibility are usually still bought.

What is vibe coding?

The term describes developing software with AI tools, where users describe requirements in natural language and the AI generates the code. That allows even small teams without a traditional engineering department to build applications.

How can software vendors protect themselves from customers building their own?

By shifting their product’s value to hard-to-copy components, offering open APIs and designing their packages so customers don’t pay for features they don’t use.

Further reading: Is SaaS dead? · Who owns the customer when the agent does the work? · What investors want to see after the SaaS correction

Sources: Cruxy survey of 300 SaaS CEOs, reported by SecurityBrief UK · OMR Reviews (2026): Die SaaSocalypse-Korrektur

Similar Posts