The technology budget for 2027 is being drawn up at an inopportune moment. The European Commission forecasts that Polish GDP growth will slow from 3.5 per cent in 2026 to 2.8 per cent in 2027, partly due to a decline in investment following the expiry of the stimulus from the National Recovery Plan. The general government deficit is set to remain high, whilst debt is expected to rise to 68.3 per cent of GDP. The economy will continue to grow, but management teams will enter the new year with greater financial discipline and less tolerance for projects without a quantifiable impact.
Technology is becoming more expensive faster than business
Global IT spending is set to rise by 13.5 per cent in 2026, to $6.31 trillion. Spending on data centre systems is rising by 55.8 per cent, and on software by 15.1 per cent. Even greater growth is evident in AI: Gartner estimates spending at $2.59 trillion in 2026 and $3.49 trillion in 2027. A large proportion of this amount will be absorbed by infrastructure built by providers and hyperscalers, but the cost will be passed on to customers in the form of cloud pricing, licences, computing power and services.
This changes the nature of the base budget. Maintaining the same application portfolio no longer means stable costs. AI features are being integrated into existing SaaS packages, billing is increasingly usage-based, and tokens and computing power are creating a new category of variable costs. Without measuring usage, a company pays for access, not for results.
The CFO does not question the technology. They question the lack of cost-effectiveness
In a Gartner survey , 75 per cent of CFOs reported an increase in technology budgets for 2026, and nearly half anticipated a rise of at least 10 per cent. At the same time, 53 per cent identified corporate IT as an area for potential cuts, whilst 42 per cent were targeting operational savings of 3–5 per cent. Capital is shifting from managing complexity to automation, data, cybersecurity and digital products.
The projects most at risk of being cut are those described in technological terms: migration, modernisation, platform, agent, new architecture. Management needs a different metric: order processing cost, month-end closing time, the percentage of issues resolved without a consultant, loss reduction or margin growth.
A project without a baseline value, an owner responsible for the outcome, and a decision date for scaling up or winding down is not an investment. It is an open-ended cost commitment.
AI requires a portfolio, not a collection of pilot projects
Deloitte reports that 79 per cent of technology leaders regard business outcomes as their top priority, yet 42 per cent of organisations report a low or zero return on investment in AI. As many as 75 per cent admit that scaling value requires a fundamental change to the operating model. The problem is not a lack of tools, but poor data, integration, security, skills and fragmented accountability.
FinOps experience shows how quickly the cost of AI ceases to be marginal. By 2026, 98 per cent of the teams surveyed are already managing their AI expenditure, compared with 31 per cent two years earlier. Many organisations are funding new implementations with savings achieved through cloud optimisation and other technology cost reductions. The era of separate discussions about the cloud, SaaS, data centres and AI is therefore coming to an end. A single technology cost portfolio is managed, linked to the product, process and business owner.
Klarna demonstrates what a data-driven investment case looks like. According to the company’s filing with the SEC, its AI assistant handled 80 per cent of customer chats and 31 million conversations in 2025. The company estimates the equivalent of over 700 full-time roles and $39 million in savings in 2024. The time taken to resolve an issue has fallen from 12 to two minutes, whilst the number of repeat enquiries has dropped by 25 per cent. These are the company’s own figures, but they combine cost, volume, quality and operational impact — precisely what is missing from most budget presentations.
Not every cost is worth optimising on a percentage basis
Cybersecurity, resilience and AI oversight do not generate direct revenue, but they have a quantifiable impact. IBM estimates the average global cost of a data breach at $4.44 million. Organisations making extensive use of AI and automation in security recorded an average cost per incident that was $1.9 million lower, whilst 63 per cent of those surveyed had no AI management policy.
From 2 August 2026, most of the provisions of the EU AI Act will already be in force, whilst the requirements for certain high-risk systems will come into effect on 2 December 2027. The budget must include model inventory, access control, monitoring, documentation and testing. Removing these items does not reduce the cost of the technology. It simply shifts it to operational, legal and reputational risks.
The IT budget for 2027 should show three figures for each major item: the full cost over a horizon of several years, the measurable business impact, and the cost of delay or cancellation. Without these, the board will make percentage-based cuts. With them, it can reallocate capital from maintaining technological debt to projects that boost margins, resilience or growth rates.
