5 minBusiness
CIOs Weigh Build vs. Buy as AI Coding Tools Reshape Enterprise Software
As AI coding tools make it easier for companies to build their own software, a software executive and former CIO outlines five questions leaders should ask before deciding whether to build in-house or buy an established platform.
Enterprise technology leaders are confronting a renewed build-versus-buy debate as AI coding tools such as Claude Code make it easier for companies to create custom applications in-house. The pitch is straightforward: instead of paying recurring subscriptions for platforms like SAP, Workday, or HubSpot, CIOs can direct their engineering teams to build replacements and keep the savings.
But a former CIO who now leads at a global software company that helps large enterprises manage workflows for HR, customer service, and finance warns that the calculation is rarely that simple. Writing from the perspective of someone who has sat in the CIO seat and now sells enterprise software, he argues that there are times to build with AI and times when buying is the obvious choice. The stakes, he notes, extend beyond budgets to the CIO's own job.
The first question is whether software development is the company's core competency. Even organizations with deep engineering talent can lose focus by building tools unrelated to their main business. One global technology company with hundreds of thousands of employees chose not to build its own billing and HR software, preferring to concentrate on innovation. For most businesses, the priority should be leaning into existing expertise rather than constructing something outside their mission.
Cost is the second consideration, and the savings pitch often begins with the subscription fees a company would eliminate. Those fees can reach millions of dollars a year at large enterprises. However, the executive says companies building their own LLM-based software solutions typically spend five to 10 times more than they would using a workflow automation platform, depending on business complexity. Initial development is only the beginning; maintenance, security updates, and keeping the system current as the business evolves drive the real expense. At one large financial services company, the chief information security officer wanted to use an LLM to rewrite a core system, a process that could take 18 to 24 months to save the equivalent of just 0.5% of the company's annual operating budget.
Security and governance form the third question. Getting software to work on day one is not the hard part; defending it on day 1,000 is. AI agents can exceed their permissions, and roughly half of organizations have seen that happen. The car rental management platform PocketOS had its production database wiped by a coding agent in nine seconds. CIOs must honestly assess whether their platform can withstand evolving threats, whether it is auditable for regulators, whether permissions can be tracked, and whether the software can comply with shifting rules.
The fourth question concerns leadership continuity. The average tenure for a CIO is about 4.5 years, while a custom platform typically takes two or more years to build and longer to mature. The person who designed the architecture and knows where every integration lives may be gone before the system runs well. One CIO described a predecessor who built a custom platform the entire organization depended on. When he left, the institutional knowledge left with him. There was no documentation, no support team, and no vendor to call, leaving senior leaders with a system they could not explain and could not afford to shut down.
Finally, leaders should ask whether they can tap into proprietary data and institutional knowledge. Building in-house makes sense when projects differentiate the business and draw on a company's «secret sauce.» A chief digital and information officer at one of the world's largest shipping and logistics companies is building with AI but is highly selective, greenlighting only projects that harness decades of internal data and institutional knowledge to optimize operations.
The executive's broader message is that AI has not eliminated the trade-offs that have always shaped enterprise software decisions. Building can create value when it sharpens a company's competitive edge, but it can also drain resources, introduce security risks, and leave organizations dependent on knowledge that walks out the door. For CIOs weighing the choice, the questions are less about what AI makes possible and more about what the business should actually do.
13
