Rising AI costs are forcing organisations towards more mature AI governance

Rising AI costs are forcing organisations towards more mature AI governance:

Why technology alone is not enough

The first wave of AI adoption was driven by enthusiasm. Today, executive teams are facing a new reality: rising AI costs, unclear return on investment, and a growing number of experiments that never evolve into sustainable business improvements.

A recent article in De Tijd highlighted how organisations worldwide are struggling with increasing AI bills while searching for the right balance between innovation and cost control. At Select Advisory, we recognise this shift. However, we believe the challenge extends far beyond rising token costs. The real question is not "How can we make AI cheaper?" but rather "How can we ensure AI delivers sustainable business value?"

Our experience shows that successful AI initiatives rarely fail because of the technology itself. They fail because organisations are not sufficiently prepared for the organisational transformation that AI requires.

From AI hype to AI reality

In De Tijd (27 June), the situation was summarised perfectly: "Many of our customers had already fixed their IT budgets, only to discover that their AI costs are now three to ten times higher than expected." This is no longer an isolated case. Organisations are rapidly experimenting with generative AI, copilots and AI agents. While an initial chatbot may have cost only a few euros per user, today's more advanced applications generate exponentially higher token consumption and increasingly unpredictable invoices. This is hardly surprising. More powerful capabilities inevitably lead to longer prompts, more sophisticated analyses, autonomous AI agents executing multiple tasks, and a growing number of employees using AI on a daily basis.

The cost per interaction is not necessarily increasing. Instead, AI is simply being used far more intensively. This is a classic example of the Jevons Paradox: when technology becomes more efficient and cheaper to use, overall consumption often grows faster than the reduction in cost per unit. Yet while much of the discussion focuses on controlling AI expenditure, we believe the real challenge lies elsewhere.

Many organisations are eager not to miss the AI opportunity and therefore move quickly into implementation without first answering some fundamental strategic questions:

• Which business processes do we genuinely want to improve? Where can AI create measurable business value? Which decisions should AI support? Who owns the output generated by AI? Which data can we actually trust?

Without clear answers, AI investments, large or small, often fail to deliver the expected benefits in either the short or long term. They fail to deliver because of weak governance, unclear ownership and insufficient organisational maturity.

Based on our experience at Select Advisory, these are some of the key recommendations we share with our clients:

1) Data quality still determines AI quality

One of the oldest principles in analytics remains just as relevant today: Garbage In, Garbage Out. AI can never outperform the quality of the data it relies upon. If poor-quality or unreliable data is used to automate business processes, AI simply makes mistakes faster, or worse, automates those mistakes at scale.

Yet many organisations continue investing heavily in AI models while their underlying data remains:

• scattered across multiple systems;

• based on inconsistent definitions;

• lacking clear ownership;

• insufficiently reliable for automated decision-making.

If these issues remain unresolved, the risk of Garbage In, Garbage Out (GIGO) becomes significant. Rather than accelerating organisational strengths, AI simply amplifies existing weaknesses.

2) AI belongs in the boardroom

Much of the public discussion still frames AI as an IT budget issue. We see it differently. AI is fundamentally a strategic leadership challenge. Every member of the executive team views AI through a different lens.

For the CEO, the focus is competitive advantage, scalability and long-term business impact. The goal is to avoid endless pilot projects or disconnected experiments that never become organisational capabilities. For the CFO, the key questions are financial: How will AI make business processes faster, more efficient or less expensive? What is the expected return? And what will the total cost ultimately be? For the CHRO, the challenge revolves around people. AI changes roles, required skills and leadership expectations. Employees need to learn how to work confidently with AI, leaders must become comfortable making data-driven decisions, and organisations need to develop entirely new capabilities.

Without these organisational changes, technology remains exactly that: technology.

3) Transparency is becoming the new success factor

One of the more interesting observations in the De Tijd article concerns organisations making AI consumption visible. This represents an important step forward. After all, you can only manage what you measure. But transparency should extend well beyond token usage. Organisations should also be able to answer questions such as:

• Where is AI currently being used? Which processes generate measurable business value? Which AI models are appropriate for which use cases? Where do the greatest risks lie? Which business cases justify further investment?

Only when organisations can answer these questions does genuine AI governance begin to emerge.

4) Less technology, more strategy

The article includes an excellent analogy: "You don't necessarily take your Ferrari to drive to the supermarket." The same principle applies to AI. Not every task requires the most powerful or the most expensive model. Yet many employees routinely use advanced AI models for relatively simple activities such as drafting emails, translating documents, summarising reports or performing straightforward analyses.

A well-designed model strategy can significantly reduce AI costs without sacrificing quality. Achieving that requires more than simply monitoring token usage after the fact. It requires governance.

Sustainable AI for process excellence

This is precisely where Select Advisory supports its clients. We do not simply help organisations implement AI. We help them make AI sustainable. Everything starts from one fundamental principle: AI only delivers value when the organisation is ready for it.

That is why we never begin with the technology itself. Instead, we start with the foundations:

• an AI & Data Readiness Scan;

• identifying business processes with the greatest value potential;

• assessing data maturity, governance and ownership;

• developing robust business cases;

• supporting organisational change;

• implementing AI in carefully phased, measurable steps.

This approach helps organisations avoid the classic trap of technology push, where technology advances faster than the organisation's ability to adopt it effectively.

AI only delivers returns when the organisation evolves too

The De Tijd article concludes with three practical recommendations:

• use AI for the right tasks;

• use it in the right way;

• choose the right model.

We fully agree. However, we believe sustainable AI requires a fourth principle: Ensure your organisation is genuinely ready to realise AI's potential. Ultimately, the greatest value does not come from cheaper tokens. It comes from better business processes, better decisions and organisations that no longer see AI as merely another technology project, but as a strategic business transformation.

Conclusion

Rising AI costs cannot simply be solved by selecting cheaper models. They are a clear signal that organisations are entering the next stage of AI maturity. Organisations that want to realise lasting value from AI must look beyond the technology itself and invest in governance, data quality, ownership and organisational change.

At Select Advisory, we believe AI only creates sustainable value when technology, data, processes and people come together within one integrated approach.

The success of AI is not determined by how much AI an organisation uses. It is determined by the quality of the organisation using it. That is the essence of sustainable AI for process excellence.