News
Agentic AI Could Cut Billions in Losses on Major Infrastructure Projects
A new EY-Parthenon report estimates that errors and rework eat up to 15 percent of the value of infrastructure projects worldwide, and that agentic AI could shrink decision processes taking weeks down to a few hours. The authors flag Poland as a country with especially high stakes given the scale of its planned investments.
Contents
Consulting firm EY-Parthenon published a report titled "The Intelligence Layer" on July 15, 2026, estimating that agentic artificial intelligence could recover part of the hundreds of billions of dollars lost each year worldwide to errors, rework and delays in major infrastructure projects. The authors single out Poland as a country where the stakes in this game are especially high.
The report's starting point is a simple observation: over this decade the world needs to spend trillions of dollars a year on infrastructure, and a significant share of that money is wasted not through bad investment decisions but through inefficiency in the delivery process itself. Late-stage design changes, delayed decisions, poor coordination between contractors and a lack of early risk detection add up to costs that, as the authors stress, could have been avoided.
Scale of the problem
The Construction Industry Institute, a research organization that has studied construction-sector efficiency for years, puts the direct cost of rework at about 5 percent of project value. EY-Parthenon goes further, also counting indirect effects: schedule slippage, idle equipment, contractual penalties and lost revenue from later project completion. Once those factors are included, the report puts the total cost of inefficiency at 10-15 percent of project value, which, against global spending in the trillions of dollars a year, translates into hundreds of billions of dollars in losses.
What the intelligence layer does
The concept described in the report, called the "intelligence layer," is not a single piece of software but a system of cooperating autonomous AI agents that monitor a project continuously. Instead of waiting for periodic reports and status meetings, the agents are meant to analyze documentation, schedules and site data in real time, catching discrepancies before they grow into costly problems.
According to EY-Parthenon, a system like this can assess the consequences of a design change across an entire chain of dependencies far faster than a human team, cutting decision-making from weeks to a few hours. On top of that comes coordination across data silos, which in large infrastructure projects typically belong to different contractors, subcontractors and institutions, as well as detection of the faintest risk signals, small deviations that raise no alarm individually but together point to a growing problem.
The report also highlights the role of so-called digital institutional memory, a record of decisions, rationales and outcomes from previous projects that AI agents can draw on instead of starting each risk assessment from scratch. In large organizations running many investments in parallel, this memory is expected to be one of the main sources of savings.
Why Poland's stakes are high
Maciej Ziomek of EY-Parthenon Poland notes that the country is now entering a period of exceptionally heavy infrastructure spending. The rebuilding and expansion of energy grids, rail and road investments, and security-related projects add up to a scale the Polish economy has never taken on simultaneously before.
Poland is facing an unprecedented wave of infrastructure investment - Maciej Ziomek, EY-Parthenon Poland partner
Ziomek stresses that at this scale, even a small percentage of inefficiency translates into very large sums in absolute terms, and a significant share of investment in Poland is carried out by the public sector, where tendering and acceptance procedures are particularly prone to delays.
What matters is not just how much we invest, but above all how effectively we deliver projects - Maciej Ziomek, EY-Parthenon Poland partner
The report does not name specific Polish projects where the intelligence layer would be deployed, but Ziomek's comment fits a broader pattern: Polish construction firms and public investment bodies have for months been signaling that administrative procedures, not execution capacity, are the main factor stretching out delivery of large contracts.
What this means for companies
For Polish companies in construction, energy and infrastructure, the report's takeaways are practical: the agentic AI described by EY-Parthenon does not replace engineers or decision-makers, but is meant to speed up the flow of information between them and flag deviations from plan earlier. In practice, that means a potential competitive edge for firms that are first to deploy such tools in managing large contracts, especially public ones, where contractual penalties and lost EU funding can be severe.
The EY-Parthenon report joins a growing number of analyses from consulting firms that in 2026 point to agentic AI as a cost-saving tool in specific, capital-intensive sectors of the economy, not just as a generic productivity booster for office work.
