Artificial intelligence is beginning to change how companies decide whether to buy software or build it themselves. A new McKinsey Global Survey found that 32% of respondents at organizations regularly using AI said their companies had decided against purchasing at least one software product or feature because they could build the functionality internally with AI-powered coding tools. The finding highlights a growing shift in enterprise technology budgets as coding agents make custom software development faster and more accessible.
The trend is particularly strong among technology companies, where 41% of respondents reported that their organizations had forgone at least one software purchase in favor of an in-house build. Healthcare followed at 39%, while professional services and energy and materials were both at 38%. Financial institutions recorded 36%. McKinsey’s survey covered 1,719 participants across nearly 100 countries between May 4 and June 8, 2026.
AI Coding Tools Are Changing The Build-Vs-Buy Decision
For decades, enterprises have generally relied on commercial software vendors for specialized applications. Buying software provided predictable functionality, maintenance, security updates and integrations without requiring companies to build large internal engineering teams.
AI coding agents are changing that calculation.
These tools can assist with planning, writing, testing and modifying software, allowing smaller teams to create applications tailored to specific internal requirements. McKinsey’s latest research indicates that this capability is beginning to affect actual procurement decisions rather than remaining limited to experimentation.
The distinction is important. The survey does not say that 32% of companies are replacing all commercial software with internally developed applications. Instead, it measures organizations that have decided against buying at least one software product or feature because AI coding tools allowed them to build the functionality internally.
Industries Most Likely To Build Instead Of Buy
| Industry | Share Forgoing At Least One Software Purchase |
|---|---|
| Technology | 41% |
| Healthcare payers and providers | 39% |
| Professional services | 38% |
| Energy and materials | 38% |
| Financial institutions | 36% |
| Media and telecommunications | 34% |
| Pharma and medical products | 33% |
| Travel and logistics | 31% |
| Consumer goods and retail | 29% |
| Engineering and construction | 28% |
| Advanced manufacturing | 27% |
| Insurance | 19% |
| Public and social sector | 17% |
| All respondents | 32% |
The figures show that the trend extends well beyond software companies. Industries with complex internal processes, proprietary data or specialized workflows are also finding reasons to develop customized applications rather than purchase generic products.
Technology Companies Lead The Shift
Technology organizations are at the forefront of the change, with 41% reporting that they had skipped at least one software purchase in favor of an internal AI-assisted build.
That is not surprising given the concentration of engineering talent, data expertise and software infrastructure within the sector. Technology companies can also have highly specialized requirements that commercial products may not address efficiently.
Healthcare’s 39% figure is notable because organizations in the sector often operate complex workflows involving patient information, compliance requirements and specialized systems. Building certain functions internally can offer greater control, although regulatory and security requirements can make enterprise software development considerably more demanding.
Professional services and energy and materials both recorded 38%, while financial institutions stood at 36%.
Larger Companies Are Scaling AI Coding Faster
McKinsey’s broader 2026 AI research shows that AI adoption and scaling are increasingly concentrated among larger enterprises.
Forty percent of respondents from organizations with annual revenue above $1 billion said their companies were scaling AI agents, compared with 22% at smaller organizations. Software coding agents are also being scaled by roughly one-third of respondents at larger enterprises.
This creates an important connection between company size and the build-versus-buy trend.
Larger enterprises typically have more data, established technology teams and larger software budgets. They therefore have both the resources and the potential economic incentive to use AI coding tools to develop specialized internal applications.
AI Adoption And Enterprise Scaling
| McKinsey 2026 Indicator | Reported Share |
|---|---|
| Organizations using AI in at least one function | 88% |
| Large enterprises scaling AI agents | 40% |
| Smaller organizations scaling AI agents | 22% |
| Organizations scaling chatbots | 47% |
| Organizations scaling AI agents | About 20% |
| Organizations scaling software coding agents | About 20% overall |
| Organizations avoiding at least one software purchase through AI coding | 32% |
The data illustrates an important difference between AI adoption and AI maturity. Although 88% of organizations report using AI in at least one business function, McKinsey says only around 1% consider themselves fully mature in AI deployment.
AI Is Not Yet Delivering Broad Financial Gains
The increase in internal software development is happening even as many companies struggle to translate AI use into measurable financial results.
McKinsey’s 2026 survey found that 37% of respondents reported some positive earnings impact from AI, roughly unchanged from the previous year. At the same time, around 80% said AI had improved their individual productivity, while about half said it helped them make better decisions.
This gap between individual productivity and company-wide financial impact is becoming one of the defining issues in enterprise AI adoption.
From Productivity To Business Impact
AI Adoption
│
├── 88% use AI in at least one function
│
▼
Employee-Level Benefits
│
├── ~80% report improved productivity
├── ~50% report better decision-making
│
▼
Enterprise-Level Outcomes
│
├── 37% report some earnings impact
└── ~1% consider themselves fully AI mature
The build-versus-buy trend could be one mechanism through which companies attempt to close that gap. Instead of using AI merely as a productivity assistant, businesses can use coding agents to redesign workflows and create software specifically around their own operations.
Why Companies May Choose In-House Software
There are several potential reasons for choosing an internal build.
First, AI can reduce the time required to create relatively simple applications. A company that previously needed months of engineering work for a customized dashboard, workflow automation or internal tool may be able to produce an initial version much faster.
Second, internally built software can be tailored to proprietary processes. Commercial software generally has to serve a broad customer base, whereas an internal application can be designed around one company’s data and workflows.
Third, companies can potentially reduce dependence on multiple SaaS vendors for narrow functions. If a business needs only a small portion of a commercial platform’s capabilities, developing that functionality internally may become economically attractive.
McKinsey has previously estimated that generative AI could cause a 2-to-4 percentage-point shift in software spending from buying toward building, equivalent to roughly $35 billion to $40 billion, although that earlier analysis covered a longer-term industry outlook rather than the 2026 survey’s actual adoption figure.
The Software Industry Faces A New Challenge
For software vendors, the trend raises questions about pricing, product differentiation and customer retention.
The risk is greatest for products that provide relatively straightforward functionality that can be reproduced using an AI coding assistant. Basic dashboards, internal workflow tools, simple data-query applications and narrow productivity features could face greater pressure as development costs fall.
More complex systems may remain harder to replace. Software deeply embedded in business processes, supported by large ecosystems or holding critical enterprise data can be more difficult to replicate.
McKinsey has previously argued that generative AI could increase software vendor switching by lowering the cost of data migration, integration development and user training.
What Could Become More Vulnerable?
| Software Category | Potential AI Pressure |
|---|---|
| Simple internal tools | High |
| Basic dashboards | High |
| Workflow automation | High |
| Ad hoc data querying | High |
| Specialized enterprise systems | Moderate |
| Systems of record | Lower relative risk |
| Highly regulated platforms | Lower, but compliance remains critical |
| Complex industry-specific software | Variable |
The implication is not that commercial software is becoming obsolete. Instead, the market may increasingly divide between software that companies can cheaply recreate and software whose data, integrations, reliability and regulatory infrastructure provide substantial barriers to replacement.
AI Coding Agents Could Reshape IT Budgets
The build-versus-buy decision could eventually affect how chief information officers allocate technology budgets.
Historically, organizations divided technology spending between software licenses, cloud infrastructure, external services and internal engineering. If AI coding tools make internal development more efficient, some spending could shift toward engineering platforms, AI models, data infrastructure and internal development teams.
That could benefit AI model providers, cloud companies and developer-tool businesses even as it puts pressure on some traditional SaaS vendors.
The change could also create a new requirement: companies will need stronger governance over internally developed AI-generated applications.
Cheap development does not eliminate the need for security testing, documentation, maintenance, access controls, compliance reviews and long-term ownership.
The Hidden Cost Of Building Software
McKinsey’s research on AI-native companies highlights a key limitation: building a customized tool may be inexpensive initially, but maintaining it can become expensive over time. The consultancy advises organizations to build what is genuinely distinctive while buying or using external tools for capabilities that do not provide strategic differentiation.
This means the build-versus-buy decision is unlikely to disappear. Instead, AI may change where companies draw the line.
A company might purchase a core enterprise platform while using AI coding agents to build specialized layers around it. That hybrid model could become increasingly common.
The Bigger Picture
McKinsey’s latest findings show that AI is beginning to influence software procurement itself, not merely how employees use existing applications. With 32% of AI-using organizations reporting that they have already skipped at least one software purchase because they could build the functionality internally, coding agents are emerging as a new factor in enterprise technology strategy. The strongest adoption is appearing in technology, healthcare, professional services, energy and materials, and financial services.
The broader significance is a potential shift in the economics of software. If AI continues to reduce development time and expand the number of employees capable of creating applications, companies could increasingly build highly customized tools around proprietary workflows and data. However, the cost of maintaining, securing and governing those applications means the future is unlikely to be purely “build instead of buy.” The more likely outcome is a hybrid software market in which companies buy core platforms while building increasingly specialized AI-enabled functionality around them.
Looking Ahead
The next phase of enterprise AI adoption will likely focus less on whether companies use AI and more on where they can generate measurable economic value from it. McKinsey’s data suggests that organizations are already scaling coding agents and using them to reconsider software procurement, but the relatively limited share reporting direct earnings benefits shows that technological capability alone is not enough. Companies will need to redesign workflows, establish governance and connect AI investments to measurable business outcomes.
For software vendors, the challenge will be to make products difficult to replace by offering capabilities that internal teams cannot easily reproduce, including proprietary data, deep integrations, reliability, security and specialized expertise. For enterprise technology leaders, the challenge will be deciding which functions create enough strategic value to justify building internally. As AI coding agents improve, that boundary is likely to keep moving, potentially making software development itself a more important part of the enterprise technology strategy.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
