Decoding the 18X Cost Surge
During its recent IT Symposium, research firm Gartner dropped a bombshell statistic that is reverberating through boardrooms and IT departments. An analyst noted that while only one in five AI projects succeeds, those successful projects now come with
a price tag 18 times higher than they did just two years ago. This isn't a simple case of rising software license fees. Instead, it points to a much deeper issue: as companies move from small-scale pilots to enterprise-wide deployment, the true costs of making artificial intelligence work effectively are coming into sharp focus. These expenses are often hidden, buried in budgets for infrastructure, data management, and specialised talent, but they are very real.
The Perilous Gap: Adoption vs. Governance
The core of the problem lies in a dangerous chasm between ambition and readiness. A recent Gartner survey found that while 71% of organisations plan to deploy agentic AI within the next year, only 21% currently have the governance capabilities to manage it. This creates a 50-percentage-point gap where risk can flourish. In the rush to innovate and stay competitive, many companies are rolling out powerful AI tools without the necessary guardrails. Experts warn this 'act now, govern later' approach can lead to millions of dollars in unexpected expenses from unmonitored employee use and unrestricted licenses. The risk is so significant that Gartner has flagged it as a 'material margin risk' for businesses.
Where the Hidden Costs Lie
Most leaders forecast AI budgets based on visible costs like subscription fees, but this is merely the tip of the iceberg. The real expenses are often lurking beneath the surface. One of the biggest drivers is token consumption. AI models process information in units called 'tokens', and every query, response, and interaction racks up costs. Gartner predicts that by 2028, the cost of AI coding agents alone could surpass the average developer's salary. Beyond tokens, significant costs include: data preparation and cleansing to make information usable for AI; the high-performance computing infrastructure needed to run models; and the extensive employee training required to use the tools effectively and responsibly.
The 'Work Slop' Productivity Drain
Another hidden cost comes from a phenomenon Gartner calls 'work slop'. This refers to the low-quality, inaccurate, or irrelevant output from poorly implemented AI that employees then have to spend time correcting. One estimate suggests 40% of workers have received AI-generated 'slop' and spent, on average, two hours a month fixing it. For a large company, this lost productivity can translate into tens of millions of dollars annually. This directly contradicts vendor claims of massive efficiency gains and highlights how, without proper oversight, AI can sometimes create more work than it saves. This inefficiency is a direct consequence of a culture that prioritises deployment speed over output quality.
From Cost Centre to Value Driver
The solution isn't to abandon AI but to adopt a more disciplined approach. Experts recommend establishing strong AI governance frameworks before scaling projects. This involves creating a centralised view of all AI-related spending and implementing cost controls, a practice often called 'AI FinOps'. Companies also need to shift their focus from technology alone to building a data-driven culture. Gartner research shows that cultural resistance, not budget, is the biggest blocker to successful AI governance. Success depends on making smarter architectural choices, selecting the right-sized model for each task, and ensuring there is clear accountability for AI-driven outcomes. Ultimately, cost governance needs to be seen as a non-negotiable part of any AI strategy, ensuring that these powerful tools deliver sustainable value rather than just a shocking bill.
















