The AI Reckoning Is Coming — And Most Organisations Are Not Ready for It

What started as experimental budget is now a significant and growing cost line — and the CFO is arriving with questions that most AI programmes cannot yet answer

Video summary: https://youtu.be/9f2Qyc5T_wg

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/the-ai-reckoning-is-coming-and-most?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

For the past two years, AI spending in most organisations has enjoyed a kind of protected status. It lived in innovation budgets, discretionary funds, and technology experiment lines that finance teams largely left alone. The assumption was that early AI investment was a necessary cost of staying relevant, and that asking too many hard questions about returns would signal a failure of strategic vision. That era is ending. Across boardrooms globally, the conversation is shifting from “are we investing in AI?” to “what exactly has it cost us, and what have we got back?” For most organisations, the honest answer to that second question is deeply uncomfortable — and the CFO has only just begun to look.

CONTEXT AND BACKGROUND

The scale of AI spending is no longer modest at any level. According to Computerworld, generative AI budgets have increased substantially year over year, yet a majority of organisations still struggle to demonstrate sustained return on investment. Early pilots often look promising, but value becomes harder to explain as systems scale, costs fluctuate, and governance expectations rise. One CIO quoted in the publication captured the problem bluntly: if the best a team can report to the CFO is that 95 per cent of employees are using AI, that is not a financial argument — it is the equivalent of saying everyone uses email.

The numbers at the infrastructure level are staggering. According to a TechCrunch article, the five largest cloud and AI providers have collectively committed to spending between 660 and 700 billion dollars on capital expenditure in 2026 alone — nearly double 2025 levels. That capital filters down to every enterprise through subscription fees, compute charges, licensing costs, and implementation services. For organisations that have embedded multiple AI tools across business units, the cumulative cost is growing faster than most finance teams realise.

INSIGHT AND ANALYSIS

What makes the AI cost problem particularly difficult to manage is that it does not behave like traditional technology spending. According to Analytics Week, token-based pricing models fluctuate based on context length, retry behaviour, and user interaction patterns. A seemingly minor change in how a prompt is structured or how frequently a tool is used can double inference costs almost overnight. This volatility makes traditional budgeting models — built around predictable compute and storage usage — largely ineffective. AI spend is also organisationally fragmented. It is buried in shared cloud accounts, scattered across team budgets, and masked by platform abstractions. What looks like innovation to a product team can appear as uncontrolled burn to a financial controller. In many cases, organisations have not realised how expensive their AI systems have become until budgets are already exceeded.

The inference dimension of this problem is accelerating. Research published by AI Ireland highlights a critical distinction that most boards have not yet internalised: AI used for day-to-day operational efficiency belongs in operating expenditure, while AI infrastructure that creates new revenue streams or durable capabilities may be better classified as capital expenditure. Most organisations have not made this distinction clearly, which means their financial reporting on AI is structurally incomplete. Around 62 per cent of organisations plan to increase AI spending this year, yet the era of blank-cheque pilots is finished. Leaders are being asked to channel investment into targeted projects with clear returns — but many lack the measurement frameworks to know which projects qualify.

The governance dimension is equally concerning. Lloydson’s research on AI infrastructure trends found that organisations are spending 40 to 60% more on AI infrastructure than they originally budgeted, driven by unexpected compute costs, data storage requirements, and the need for redundancy and reliability at scale. For a large organisation running production AI systems, monthly infrastructure costs can easily run into millions of rands — yet many leadership teams have no unified view of what that spend totals, which business units are responsible for it, or what measurable outcomes it is producing.

IMPLICATIONS

For CFOs and finance teams, the immediate priority is visibility. Before any conversation about ROI can be had, organisations need a single, consolidated view of what AI is actually costing them across all tools, platforms, teams, and usage patterns. Without that foundation, every board conversation about AI investment is built on incomplete information. The discipline now emerging globally — sometimes called FinOps for AI — applies the same accountability principles to artificial intelligence that cloud FinOps brought to cloud computing a decade ago. South African organisations would do well to adopt it now, before the costs compound further.

For boards and executive teams, the question to ask is not whether AI is delivering value in isolated pockets — it almost certainly is, somewhere. The question is whether the total cost of the organisation’s AI portfolio is proportionate to the total value it is generating, and whether that case can be made clearly and credibly under scrutiny. Most AI programmes cannot yet answer that question. Those who cannot should treat it as urgent.

For AI and technology leaders specifically, the window for building the right cost governance structures is narrowing. Boards that have been patient with AI experimentation are beginning to apply the same financial rigour to AI that they apply to every other category of significant spend. The leaders who get ahead of this conversation — who arrive at the board table with a clear cost map, a credible ROI framework, and an honest account of what is working and what is not — will be far better positioned than those who are still treating AI spend as exempt from normal financial accountability.

CLOSING TAKEAWAY

The free pass on AI spending is over. What began as an investment in future relevance is becoming a significant, growing, and largely ungoverned cost line inside organisations that have neither the frameworks nor the governance structures to manage it with the rigour it now demands. This is not a reason to slow AI adoption — the competitive and operational case for AI remains compelling. It is a reason to grow up about how that adoption is funded, tracked, and justified. The CFO is coming. The board is asking questions. And for organisations that have been spending on AI without a clear account of what it is returning, the reckoning is no longer a future risk. It has arrived.

Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net