A recent McKinsey report has put a number on something software vendors have been quietly worried about for a while: nearly a third of surveyed companies decided against purchasing at least one software product or feature because they realized they could build the same functionality themselves using AI-powered coding tools.
Why this is a bigger deal than it sounds
Software licensing has traditionally been a fairly stable, recurring line item in corporate budgets — the classic "buy, don't build" logic held because building in-house was slow, expensive, and required specialized engineers. AI coding agents chip away at exactly that calculation. If a team can generate a working internal tool in days instead of months, the cost-benefit math for licensing an off-the-shelf product changes meaningfully, especially for narrower, less complex features.
Who this actually threatens
The pressure isn't evenly distributed. Broad, complex platforms with deep integrations and compliance requirements are much harder to casually replace with an internally built tool. It's narrower, well-defined features — the kind of thing a single product manager might describe in a page of requirements — where AI-assisted internal builds are becoming a genuine substitute for buying.
What comes next for vendors
The report suggests software companies are increasingly under pressure to build capabilities that are genuinely hard to replicate with a quick internal AI-assisted build — deep domain expertise, complex integrations, or compliance and security work that a fast internal tool would struggle to match. Vendors that lean mainly on convenience rather than defensible complexity look the most exposed to this shift.
Reporting on this story also appeared at AI Agent Store, which has more technical detail if you want to go deeper.