McKinsey published a study in March 2026 that, on the surface, is about budget allocation. But anyone who reads past the figures encounters a more uncomfortable message: most organisations spend their technology budgets in ways that structurally undermine AI adoption. Not by investing too little, but by investing in the wrong way.
What was studied
The paper “Recalibrating technology budgets for the AI era” draws on a survey of 17 large companies in Australia, Europe and North America covering their IT spending in 2024 and plans for 2025. The authors, Leandro Santos, Thomas Elsner and Ishaan Sharma, mapped the relationship between two types of spending: “run”, the costs of keeping existing systems operational, and “change”, the investments in renewal, modernisation and new capabilities. AI budgets are not a separate category in this analysis. They are a component of both.
The central finding: AI is absorbing up to a third of large organisations’ change budgets, but barely reduces run costs. In many companies, operational costs are actually rising, because AI adds new systems on top of existing infrastructure without anything being decommissioned.
Four types, one pattern
McKinsey identifies four IT archetypes based on the ratio between run and change investment.
“Deliberate modernizers” allocate at least a third of their budget to change (37%), keep run costs structurally low and operate on standardised platforms. Their technical debt is manageable. They are the benchmark.
“Strained transformers” also invest heavily in change (34%), but build on top of legacy infrastructure. Run costs remain high (66%) as a result, and returns on change investment risk stagnating. They are moving, but carrying too much weight.
“Lean operators” keep everything tight: low run, but also low change (13% change, 87% run). Stable in the short term, but vulnerable to innovation stagnation.
“Heavy IT sustainers” spend the largest share on run (80%), with limited room for change (20%). This type emerges from structural complexity, or following a major modernisation wave in which the environment is deliberately kept stable.
What deliberate modernizers do differently
The difference lies not in the size of the budget, but in three specific choices.
First, they keep their run infrastructure costs at least 20% lower than comparable organisations. This is not accidental, but the result of a deliberate decision to decommission platforms, consolidate services and systematically eliminate technical debt.
Second, 57% of their application budget goes to change, compared with 21 to 47% at the other archetypes. They make room by replacing something, not just by adding something.
Third, and this is the most striking finding, they invest 16% of their total IT budget in internal personnel for change activities. That is 1.5 to 4 times more than the other groups. Talent turns out to be the real bottleneck.
Why this matters for your organisation
This study focuses on large multinationals, but the pattern is recognisable in mid-sized organisations as well. Many organisations that are deploying AI today are, in effect, “strained transformers”: they add AI tools to existing workflows without replacing or simplifying anything. The result is an IT environment that is simultaneously more modern and more complex.
Five lessons for those who want to draw the right conclusions:
- Your run burden defines your room to manoeuvre. Organisations that structurally overspend on keeping existing systems running have too little left over for renewal. That is not a budget problem. It is an architecture problem. Those who do not actively manage their run costs are financing stagnation with growth money.
- Adding AI is not the same as modernising. The most common mistake in AI adoption is stacking new tools on top of an unchanged technical foundation. It costs money, increases complexity and yields no structural advantage. Every AI investment should also trigger a question: what are we decommissioning to make room for this?
- Shared platforms are not a cost item, but a lever. Deliberate modernizers invest deliberately in data platforms and standardised infrastructure, not because it is cheap, but because it makes future investments faster and less expensive. An investment that simplifies the next investment is more strategic than one that only justifies itself.
- Talent is the bottleneck nobody wants to name. McKinsey is explicit: internal capacity for change is the sharpest distinction between the best-performing organisations and the rest. External vendors can build systems. They do not build organisational capability. The knowledge your people develop does not disappear when a contract ends.
- Simplification is a strategic choice, not a cost-cutting measure. Using AI as a catalyst for simplification, as McKinsey puts it, means concretely: use AI to eliminate processes, not to automate them. Organisations that use AI to manage complexity have not understood the problem. Organisations that use AI to do less build more sustainably than those that use AI to do more.
The real lesson
McKinsey wrote this study for CIOs of large enterprises. But the structural tension they describe applies equally to a company of 200 employees as to a multinational. The ratio between what an organisation spends on keeping things running and what it spends on moving forward is a direct indicator of its strategic room to manoeuvre.
Ultimately, it is not about the size of the AI budget. It is about whether your organisation has the space and the capacity to genuinely change, or whether it keeps building new layers on a foundation it does not dare to touch.
Take an honest look at your own situation: what proportion of your technology budget goes towards preserving what exists, and what proportion towards what you want to become?