Enterprises have begun replacing sections of commercial software with custom-built agents, a trend that signals a fundamental change in how large organizations approach technology acquisition and internal development. According to a recent CIO Dive report on McKinsey research, companies are deploying these autonomous systems at an accelerating pace, often choosing to construct solutions internally rather than continue licensing features from vendors. The shift reflects both growing confidence in artificial intelligence capabilities and a desire to reduce recurring software expenses that have ballooned over the past decade.
Large enterprises lead this movement by a noticeable margin. McKinsey’s 2026 State of AI survey, which drew 1,719 responses across 97 countries, found that the build-instead-of-buy decision is most common among AI high performers. Larger organizations have several practical advantages. Bigger companies maintain dedicated data science teams, possess richer internal datasets for training, and can absorb the initial costs of experimentation more easily. They also face higher absolute software budgets, which makes even modest reductions in licensing fees translate into significant savings. Smaller businesses, by contrast, often lack the specialized talent or the volume of proprietary data needed to train effective agents, leaving them more dependent on off-the-shelf products.
Nearly one-third of respondents, 32 percent, said their organizations had decided against purchasing at least one software product or feature because they could build the functionality in-house using agentic coding tools. Nearly half of AI high performers reported doing so, compared with 31 percent of everyone else. The unit is modest: a company that built one reporting dashboard instead of buying a license counts the same as one that replaced a larger platform. Still, McKinsey said the finding could be a sign that AI is beginning to reshape how technology budgets are allocated.
Productivity gains appear across departments. Customer service teams use agents to manage initial triage, gather relevant background information, and suggest responses that human agents can approve or refine. Finance departments have automated invoice matching and exception handling, tasks that once consumed entire teams of analysts. Software development groups use agents to generate boilerplate code, review pull requests for common errors, and even draft documentation, shortening release cycles. In McKinsey’s survey, 80 percent of respondents said AI improved their individual productivity.
Translating productivity into firm-wide cost savings has proven more complicated. McKinsey found that 37 percent of organizations attribute any EBIT impact to AI, essentially unchanged from a year earlier, and one in five say AI operating costs, including tokens, already limit their use. Part of the reason lies in transition costs. Building reliable agents requires upfront investment in infrastructure, model training, ongoing monitoring, and human oversight. Many organizations still run parallel systems during testing periods, which temporarily increases rather than decreases spending. Additionally, some savings appear in non-obvious places. When an agent eliminates the need for a particular software module, the organization may still pay for the broader platform contract that bundles multiple features. Negotiating those contracts downward takes time and procurement expertise that not every company has applied yet.
The technical foundation for this trend rests on advances in large language models and orchestration frameworks that allow agents to plan, use tools, and maintain memory across extended interactions. Modern agent architectures can break complex tasks into subtasks, select appropriate software interfaces, and recover from errors without constant human intervention. Enterprises with mature cloud environments and strong API layers find it easier to embed these agents into existing processes. Those still operating on legacy systems face steeper integration challenges, which helps explain why certain industries move faster than others. McKinsey found the build-instead-of-buy decision most common in technology and healthcare, followed by professional services and energy.
Security and compliance considerations shape deployment decisions in important ways. Many companies initially restricted agents to internal, low-risk processes such as document summarization or meeting transcription. As confidence grew, they expanded into areas involving customer data, but only after implementing strict guardrails. These include human-in-the-loop approval steps for sensitive actions, detailed audit logs of every agent decision, and sandboxed execution environments that limit external access. The CIO Dive article drawing on McKinsey findings highlights the push to build in-house; McKinsey’s report adds that high performers are twice as likely to have leaders visibly committed to AI and tend to redesign workflows around it rather than insert AI into existing ones.
Talent remains a limiting factor. Demand for engineers who understand both traditional software development and modern AI orchestration exceeds supply. Companies address this gap through several approaches. Some create centers of excellence that build reusable agent components for business units to adapt. Others partner with specialized consultancies to accelerate initial deployments while training internal staff. A growing number hire product managers specifically tasked with defining agent requirements and measuring business outcomes, recognizing that success depends as much on process design as on model performance.
Vendor responses to this trend vary. Established software providers have started offering platforms that simplify agent creation within their own ecosystems, hoping to keep customers inside familiar boundaries. Others have introduced marketplaces where companies can purchase pre-trained agents for common functions. Still, the overall direction suggests a partial unbundling of monolithic software suites. Instead of buying comprehensive platforms, enterprises increasingly mix and match internal agents with best-of-breed point solutions, paying only for what they cannot or choose not to build themselves.
The shift carries implications for software pricing models. Vendors that rely on seat-based or feature-based licensing may need to rethink how they charge for value. Some observers expect a move toward consumption-based pricing that aligns more closely with actual usage once agents begin handling tasks previously performed by human users of the software. Others anticipate hybrid models where vendors provide foundational models and orchestration layers while customers supply domain-specific data and business logic.
Challenges persist. Agent reliability can falter in edge cases, leading to errors that require human correction and sometimes erode trust. Maintaining agents over time demands continuous attention as underlying models update and business processes evolve. Data quality issues that once affected reporting now directly impact agent behavior, forcing organizations to invest more in data stewardship than many anticipated. These realities suggest that while agents reduce dependence on certain purchased features, they create new categories of operational responsibility.
Looking forward, the McKinsey data points toward continued expansion: 44 percent of respondents now report AI scaling across the enterprise, up from 38 percent. The focus is shifting from simple automation to more sophisticated applications that combine multiple data sources and coordinate across departments. For example, some organizations now experiment with agents that manage entire procurement processes from requisition through payment, interacting with both internal approval systems and external supplier portals. Others explore agents that monitor regulatory changes and automatically adjust compliance workflows.
The competitive dynamics also evolve. Companies that master internal agent development gain advantages in speed, customization, and cost structure. They can adapt faster to market changes because they control the underlying logic rather than waiting for vendor roadmaps. This capability may widen the performance gap between large organizations that can afford to invest and smaller ones that cannot. Over time, however, tools and frameworks may become more accessible, allowing mid-sized firms to participate more fully in the trend.
CIOs face a strategic choice. They must balance the allure of rapid productivity gains against the need for sustainable cost management and technology governance. The most successful organizations appear to treat agents not as replacements for every software purchase but as strategic assets that complement and sometimes displace commercial tools. They invest in platforms that make agent creation repeatable, establish clear metrics for success, and maintain transparency with both employees and customers about when and how agents participate in business processes.
The movement toward in-house agents represents a maturation of artificial intelligence from experimental technology to practical business infrastructure. Enterprises no longer ask whether agents can perform useful work; they ask how quickly and broadly those capabilities can be applied. The CIO Dive coverage of the McKinsey study makes clear that this transition is well underway in many large companies, with measurable effects on productivity and early effects on software purchasing. As the technology improves and organizational experience grows, the practice of building rather than buying certain capabilities seems likely to spread further, reshaping the relationship between enterprises and the software industry that serves them.
WebProNews is an iEntry Publication