AI investments: why the big money-making machine has yet to materialise

Many corporations have invested in AI – but the big money-making machine has yet to materialise.

Now comes the painful accounting.

The enthusiasm for artificial intelligence is beginning to fade in boardrooms around the world. After years of promises and PowerPoint visions, an uncomfortable question is emerging: does any of this make financial sense, or are businesses currently paying mainly to learn expensive lessons?

Billions invested in AI – and the financial review shocks executives

An international PwC survey of 4,454 executives across 95 countries paints a far more sober picture than the AI hype suggests. Many businesses have shifted budgets aggressively towards AI, built teams, bought licences and hired consultants – yet little of this is visible in their profit and loss accounts.

56 per cent of the senior executives surveyed say that using AI has neither increased revenue nor reduced costs.

In other words, the financial impact is currently simply neutral for more than one in every two companies – despite substantial spending on infrastructure, cloud services, data preparation and specialist staff. The expected return remains a long way off.

A smaller group does at least report positive outcomes: just under 30 per cent say AI has increased their revenue. However, the ideal result – higher revenue alongside lower costs – is still unusual. Only around 12 per cent of businesses have achieved that combination so far.

The great AI mirage: expectations versus reality

This is where the optimism of recent years collides with hard reality. In many annual reports, AI initiatives were presented as the key to new markets, radical efficiency and automated operations. The technology was already treated as a given in strategic presentations, while budgets were raised again to ensure companies did not “fall behind”.

In practice, the picture is often different: many AI projects remain stuck at pilot stage, operate separately from the core business and never make it into day-to-day production use. The result is spending on tools, consultancy and internal resources, but barely any measurable added value.

• High upfront investment in infrastructure and software
• Time-consuming data preparation that was rarely planned for
• Pilot projects with no clear route into normal operations
• Unclear metrics for measuring success

For many executives, this is an uncomfortable reality check. They must explain to their supervisory boards why the promised leap in productivity has still not arrived.

AI is not “plug and play” – and that is precisely what is underestimated

One central misconception is that many companies treat AI as a new tool that can simply be purchased, switched on and immediately deliver benefits. That is not how the technology works.

AI does not behave like a mouse that you connect and can use immediately. It requires changes across the entire company.

To deploy AI effectively, businesses need to rethink their processes. Data must be structured, clean and accessible. Responsibilities, workflows and often even business models have to change. That takes time, money and patience.

Why so many AI projects reach a dead end

Many organisations lack a clear plan for embedding AI in processes that genuinely create value. Instead, numerous small, isolated projects emerge: a chatbot here, a forecasting prototype there, an internal assistant in a laboratory area. They may look good in presentations, but they are far removed from measurable gains worth millions.

An MIT report points to the same issue: according to the study, 95 per cent of attempts to introduce generative AI into companies have not yet delivered a noticeable increase in revenue. The technology also brings familiar problems:

• Hallucinations: AI systems invent facts or provide incorrect figures that go unnoticed when controls are lacking.
• Limited real-world usability: Tasks that sound simple fail because of minor details or specific rules.
• Data security: Confidential information ends up in systems whose internal workings are often not transparent.

When AI replaces employees – and everything goes wrong

Some companies took a particularly radical approach: they dismissed large parts of their workforce and replaced work with AI solutions. In presentations, this appeared to be a bold efficiency drive; staffing costs fell in the short term, making it look like a success on paper.

However, the real-world test was harsh. Service quality deteriorated, customers complained and internal operations began to stall. Some businesses had to reverse course after a short time, recruit new employees and adjust their strategy. The promised efficiency gain became an expensive experiment.

AI is currently rarely suitable as a complete replacement for people, but rather as a tool that supports employees.

Some companies underestimated exactly this distinction. Those focused solely on rapid savings risk not only wasting money, but often damaging the trust of customers and employees as well.

Why AI investment continues to rise

Despite all the disillusionment, there is no sign of AI spending being halted. Many senior executives view the current period as a necessary learning curve. PwC in particular expects a decisive milestone for AI in the corporate context to be reached around 2026.

The pressure is immense. No board wants to be seen as the one that missed the opportunity. AI is regarded as a ticket to attracting talent, impressing investors and signalling innovative strength. In many sectors, the attitude is: it is better to invest now and make mistakes than to start too late.

Corporate attitude	Typical consequence
Fear of missing the trend	Rapid pilot projects without a clear strategy
Expectation of immediate savings	Poor decisions on workforce reductions
Pressure from investors and supervisory boards	Major announcements, thin results
Belief in long-term potential	Willingness to accept short-term losses

What companies must change to make AI pay off

Businesses that want to move beyond the costly hype phase need a different approach. Crucially, AI must not remain a prestige project for the IT department; it needs to be linked directly to core measures such as revenue, margin, customer satisfaction or lead times.

Three levers for genuine value

• Clear business goals: Rather than saying “we are doing AI now”, companies need specific targets, such as fewer complaints, faster quotation processes or more accurate demand forecasts.
• Integration into key processes: AI must be embedded in actual value creation – in sales, manufacturing, logistics and service – rather than only in laboratories or innovation departments.
• Continuous monitoring: Results need ongoing review. If a model does not produce measurable improvements, it should be adapted or stopped.

Companies that take these steps seriously are already seeing visible effects. They automate parts of documentation, support case handlers, prioritise enquiries more intelligently or improve industrial maintenance processes.

Risks many underestimate – and opportunities that are real

The biggest current risks include not only technical errors or hallucinations, but also legal and organisational consequences. When confidential contract data passes through external AI services, questions arise over compliance, liability and data protection. Incorrect information given to customers can, in the worst case, lead to legal consequences.

Conversely, the opportunities lie where AI complements rather than replaces human work: summarising information, carrying out routine analyses, sorting enquiries and suggesting options. Used correctly, it can ease employees’ workloads rather than displace them.

For many businesses, the defining task of the coming years will be to find this middle ground: moving away from costly hype towards a sober, measurable use of AI. Those that succeed have the opportunity to turn today’s disillusionment into a genuine competitive advantage – instead of merely paying heavily to ride the next wave of technology.