The rise of AI-powered coding agents is beginning to change one of the most established practices in corporate technology: buying software.
Nearly one-third of respondents—32%—said their organisations had decided against purchasing at least one software product or feature because they were able to build the functionality internally using agentic coding tools, according to McKinsey’s 2026 global survey on AI.
The trend is particularly strong in technology, where 41% of respondents reported that their organisations had chosen in-house development over buying additional software. The figure stood at 39% among healthcare payers and providers, while professional services and energy and materials were both at 38%.
The development has potentially significant implications for the enterprise software industry.
For decades, companies have relied on packaged software to solve specific business problems, from workflow management to analytics and internal productivity. AI coding agents are lowering the barriers to creating customised applications, allowing internal technology teams to build functionality that previously may have required an external software vendor.
McKinsey said the shift is part of a broader change in how larger organisations approach technology. Companies are increasingly asking what their own technology teams can build rather than automatically turning to external providers.
The trend does not mean companies will stop buying software altogether. Instead, it could lead to a more selective “build versus buy” decision.
The economics may be particularly attractive when a company needs highly customised functionality that packaged software cannot easily provide. Coding agents can potentially reduce development time and allow smaller internal teams to create applications tailored to a company’s specific requirements.
The trend is already stronger among AI high performers. Nearly half of these companies said they had decided against buying one or more software products or features because they could build them internally using coding agents, compared with 31% of other respondents.
For software vendors, the challenge may therefore be shifting from simply selling applications to providing capabilities that companies cannot easily reproduce themselves.
