Build, buy or blend? The real economics of AI-enabled software
As AI makes software development faster and more accessible, businesses face a question that could reshape the SaaS market: should they build proprietary AI-enabled software, buy an established solution or combine the two?
According to McKinsey’s report, The State of AI in 2026: On the Road to ROI, 32% of surveyed organisations had decided against purchasing at least one software product or feature because they believed they could build the capability internally using agentic coding tools.
For investors, this raises questions about the long-term defensibility of software businesses. But it also creates a risk for customers, the apparent ease of building can obscure the true cost and responsibility of owning AI software.
The true cost of building
Developing a working AI application is very different from operating it at scale. Deloitte’s 2026 report, Is Your Cost Model Ready for the Economics of AI, highlights how expenditure can become increasingly unpredictable as usage grows, with costs varying by workload complexity and model selection.
Current token prices can create a false sense of affordability because AI labs are offering them at a discount. The long-term cost of bringing software in-house and relying fully on AI remains unknown, making investors cautious. Gartner forecasts inference costs per workflow could rise more than fivefold by 2028, excluding infrastructure, security and maintenance.
Ownership, risk and value
For mission-critical functions, ownership means accepting permanent responsibility for model performance, governance, scalability and regulatory compliance. The challenge is not simply getting an application to work; it is keeping it up to date as AI models, customer demand or regulations change.
Building makes most sense where the software sits at the heart of a company’s competitive advantage. For example, where it draws on proprietary data, is embedded in a mission-critical workflow or enables an operating model that competitors cannot easily reproduce. Buying is often more rational for horizontal or increasingly commoditised capabilities, where a specialist provider can spread the cost of performance, security, compliance and continuous maintenance across a wider customer base.
ICONIQ’s 2026 State of AI report found that 49% of surveyed AI companies identified application-level innovation, including workflows, integration and data, as a primary source of differentiation, compared with 14% relying on proprietary model development.
For many businesses, the answer will be a blend. Buy the underlying technology, but own the proprietary data, workflow and customer experience that create differentiation. Regulation, mission-critical processes and operational complexity can reinforce that moat.
What should investors prioritise?
The strongest software businesses will not necessarily be those developing the greatest volume of AI internally. They will be those making commercially disciplined decisions about where ownership adds value and where external technology is the better choice. For investors, the fundamentals remain unchanged. Software that has a durable advantage built around proprietary data and mission-critical workflows, supported by management teams that understand AI and can deploy it effectively to capture market share and drive growth, will succeed.