Nearly a third of companies have decided against buying a piece of software because agentic coding tools now let them build a version of it themselves in-house, according to McKinsey's latest global survey on corporate AI use, published last month and still circulating widely among enterprise technology buyers this week.
The survey, which polled 1,719 respondents across 97 countries between May and June, found that 32% of organizations have skipped at least one software purchase in favor of building the equivalent internally with AI coding agents. The shift is most pronounced in the technology sector, where 41% of respondents reported passing on a purchase to build instead, followed by healthcare payers and providers at 39% and professional services and energy at 38%.
The trend concentrates further among what McKinsey calls "high performers" — the 6% of organizations that attribute at least 5% of earnings before interest and taxes to AI. Nearly half of that group has skipped a software purchase to build in-house, compared with 31% of other companies, according to the McKinsey report.
Leaders are asking what their organisations need to build AI tools themselves.
Lieven Van der Veken, partner, McKinsey
The survey also found large companies scaling agent use more broadly: 40% of organizations with more than $1 billion in revenue now report deploying AI agents in at least one business function, up from 27% a year earlier, and the share of companies using AI in three or more functions rose from 51% to 56%.
But the report points to a gap between ambition and results. Software built internally with agentic tools succeeds roughly a third of the time, McKinsey found, compared with about two-thirds for tools purchased from outside vendors. Separate research reinforces the caution: Gartner projects more than 40% of agentic AI projects will be scrapped before the end of 2027, and Deloitte has found only about 11% of enterprises have moved agent projects into full production.
For enterprise software vendors, the build-versus-buy shift complicates a pitch already under pressure from AI-driven productivity claims: convincing a customer to buy a product now means competing not just with rival vendors, but with the customer's own engineers armed with coding agents. Analysts caution that companies tallying the savings from building in-house should weigh the ongoing cost of maintaining custom software against its lower success rate, rather than counting only the purchase price avoided.
