The reckoning over AI cost and value has begun

Agentic coding refers to the use of AI systems that can autonomously generate, test, refine and deploy software with increasing levels of independence, enabling organizations to build applications much faster than traditional development approaches. As a result, organizations are creating everything from team-specific workflow and productivity tools to experimental applications, enhancements or replacements for existing enterprise software, and niche solutions that previously would have been too costly or time-consuming to develop. In the EY survey, 91% of senior leaders at organizations investing in AI view in-house, AI-built software development as a critical way to de-risk dependence on external software providers.

While the share of organizations focusing AI investments on off-the-shelf AI solutions has grown as leaders look for quick ways to reduce costs and accelerate adoption, 76% of senior leaders say traditional software products no longer meet their organization's specific needs. Organizations may be increasingly dissatisfied with off-the-shelf software because standardized, buy-and-configure solutions are often unable to address the unique workflows, decisions, data environments and customer interactions that drive value in the AI era. This growing frustration may reflect a broader recognition that AI creates the most value when it is closely aligned to a company’s distinct business context rather than embedded in predefined operating models.

Companies should view the build-vs.-buy decision less as a binary choice and more as a question of where AI can create differentiated value. Off-the-shelf solutions remain attractive for quickly deploying common capabilities. However, leaders increasingly recognize that standardized software often fails to address the unique workflows, data, decision processes and customer interactions that differentiate their businesses. At the same time, advances in agentic coding are making it faster and more economical to develop in-house applications, with many organizations already using AI to build tailored workflow tools, augment existing software and create solutions that previously would not have been practical to develop. As a result, the most effective strategy is often not purely building or buying, but orchestrating: combining purchased software, proprietary AI-built applications, models, data and workflows into a unified system. 

AI governance should be viewed as an accelerator rather than a barrier because it enables organizations to scale AI confidently, responsibly and efficiently. As AI becomes embedded in critical business processes, governance provides the guardrails that allow leaders to move faster without increasing unacceptable levels of risk. Rather than being a compliance exercise applied after deployment, effective governance is built into AI workflows from the outset, helping organizations manage security, regulatory requirements, and model risk and operational controls while reducing the need for costly rework later. As AI-powered software development and agentic workflows become more common, governance becomes a key enabler of faster delivery, broader adoption and stronger business outcomes — not an obstacle to them.

In enterprise AI, orchestration refers to the coordination and integration of models, data, workflows, governance processes and people into a unified system that delivers business outcomes. Rather than viewing AI as a collection of separate tools or making a simple build-vs.-buy decision, orchestration connects purchased software, internally developed AI applications and business processes so they work together seamlessly. This approach matters because many organizations end up with a mix of rented, assembled, adapted and inherited technologies that fail to function as a coherent whole. Orchestration layers help assemble models, pipelines and services into end-to-end workflows while adapting them with proprietary business logic when generic software does not meet the realities of the organization.

AI investment and employee training often lag behind ambition because many organizations have not yet translated their enthusiasm for AI into a clear, enterprise-wide vision of what they are trying to achieve and how they will create value. Leaders frequently express bold aspirations for AI, but struggle to define the roadmap, align competing priorities and build a compelling narrative that helps employees understand the purpose and impact of the transformation. As a result, organizations can find themselves engaged in a flurry of AI activity — pilots, experiments and technology evaluations — without making meaningful progress toward scaled adoption. The challenge is compounded by AI’s dual nature: it can dramatically simplify work while simultaneously requiring organizations to rethink processes, operating models and skills.