Build > Buy? McKinsey Says 32% of Companies Now Build Their Own Software with AI Agents!

According to McKinsey’s newly released “State of AI 2026” report, 32% of companies decided this year to skip buying new SaaS software products and build in-house with Agentic Coding Tools instead. In the tech industry, that figure is as high as 41%.

Instead of paying hundreds of thousands of dollars a year in license fees and being stuck with a vendor’s rigid workflow, a company can have its dev team open Cursor or Claude Code, and within a sprint or even a weekend they may have a highly customised internal system that fits their pain points perfectly.

When the cost and barrier of writing code drop dramatically, many non-core SaaS products are easily replaced by internal AI agents. Companies find that by building their own AI-powered workflow, they can achieve full hyper-personalization.

Day 2 Operation is where the real story is: using AI to write a demo is fast, but upstream API schema changes, database migrations and security updates are the killers. In the past, paying for SaaS meant paying the vendor to absorb these maintenance costs (and to take the blame when things break). Now, if you build it yourself, you have to swallow contract drift and technical debt on your own.

Apart from mission-critical or compliance-related systems (such as ERP or heavily regulated systems), where “Buy” is still the safer choice, a lot of workflow automation can genuinely be “Build”.