The cost of software is falling: The cost of poor decisions isn’t

AI is making software dramatically cheaper to build. That is not the interesting part. The interesting part is what happens next; because falling software costs do not automatically improve the decisions that determine whether a business grows, retains its best customers, misprices risk, or fails to see a competitive shift coming. 

This has triggered a predictable debate across the technology industry. If software becomes dramatically cheaper to produce, where does value come from next? In my conversations with CIOs and business leaders, that question has already been answered. The discussion has moved beyond software productivity. The focus is on performance: how AI can accelerate growth, improve margins, strengthen customer relationships, reduce risk, and enable better decisions across the organisation. 

It is, ultimately, the only question that has ever mattered. 

The constraint was never the software 

Enterprises do not invest in technology for software’s sake. They invest to grow revenue, improve margins, strengthen customer relationships, and manage risk. Software has always been an enabler of those outcomes, not the source of them. 

Making it faster and cheaper to produce does not change the underlying equation. An organisation can generate code in hours and still make poor pricing decisions. It can deploy AI agents across every workflow and still allocate capital inefficiently. It can automate its back office and still struggle to become more competitive. The constraint was never the software. It was the quality of the decisions the software was meant to support. 

Most enterprises already possess enormous amounts of data, operational knowledge, and institutional experience. AI can now analyse that information at unprecedented speed. Yet better insights do not automatically translate into better decisions. Knowing what might happen is useful. Helping organisations determine what to do next and embedding those decisions into day-to-day operations is where measurable value is created. That is the problem Enterprise Agency is designed to solve: an organisation’s capacity to apply AI to governed decisions and embed those decisions into execution, not just generating recommendations but acting on them reliably, with accountability for the outcomes. 

Intelligence without context 

Critical business decisions are rarely made in isolation. Business objectives, regulatory requirements, customer commitments, operational constraints, prior decisions, and economic realities all shape how organisations act. A recommendation may be technically accurate, but without this operating context, it can easily become commercially irrelevant, or worse, actively misleading. 

As AI becomes embedded in decisions about revenue, supply chains, customer experience, risk, and compliance, organisations need more than explainability. They need a traceable record: from the assumptions that informed a recommendation, to the approval that authorised an action, to the outcome that resulted. Without that chain, AI-influenced decisions are difficult to defend, audit, or improve. In regulated industries, that is not an edge case. It is the baseline requirement. 

A different measure of success 

This shift has significant implications for how technology is bought, sold, and evaluated. For decades, success was measured by whether systems were delivered and operated effectively. The cost of building and running enterprise software was the constraint that defined the value equation, and that constraint is collapsing. 

As the cost of creating software continues to fall, value will increasingly shift to helping enterprises apply technology to outcomes that matter: revenue growth, margin improvement, working capital, resilience, competitive position. Technology providers will be judged not by what they built, but by what their clients can do because of it. Commercial success will be tied to business performance, not delivery effort. The firms that embrace that accountability and take genuine responsibility for the outcomes they help create will define the next era of the industry. 

The human roles that matter most in this environment will not be those that build or run software. They will be the decision architects, domain specialists, and governance leads who ensure that AI-driven actions are grounded in business judgment, institutional knowledge, and clear accountability. As routine engineering tasks become increasingly automated, that expertise becomes more valuable, not less. 

The next phase of AI adoption will not be defined simply by better models or cheaper software. It will be defined by how effectively organisations translate intelligence into action, and action into measurable outcomes. 

(The author is Chief Executive Officer and Managing Director, Mphasis. Views are personal.)