Poland is entering a period of record-breaking infrastructure investment. In 2026, nearly 100 billion zlotys is set to be allocated to transport infrastructure. The national investment gap is estimated at around 380 billion zlotys. Globally, the scale of the problem is even more stark. According to EY-Parthenon, the global infrastructure funding shortfall could reach USD 64 trillion, and by 2050, investments worth nearly USD 140 trillion will be needed.
The figures are impressive. The problem is that the infrastructure sector has not been suffering solely from a lack of funds for years. It is also suffering from a costly inability to make decisions in a timely manner.
With a budget of around 100 billion zlotys, every percentage point of inefficiency amounts to one billion zlotys. A billion that could be lost in amendments, delays, claims, design changes and meetings during which a dozen or so people try to establish which version of the document is the current one.
This isn’t a technological problem. It’s a management problem.
Digitisation was supposed to help. Often, however, it merely digitised the chaos
It’s hardly a revelation to say that the infrastructure sector no longer works with graph paper. Companies use BIM, cost management systems, digital schedules, documentation platforms, sensors and analytical tools. The data is there. Sometimes there’s even too much of it.
Poland’s estimated investment gap in areas including, amongst others, transport, energy and digitalisation, stands at 380 billion zlotys.
The problem is that the schedule is stored in one system, the costs in another, the project documentation in a third, whilst information on the actual progress of the work is held in the site manager’s head and in a spreadsheet sent by email once a week.
Each of these tools may work correctly. The whole thing may still work terribly.
In Poland, awareness of BIM stands at around 60 per cent, whilst roughly one-third of the respondents from the architecture and construction sector report using it. Despite further implementations, the sector’s productivity has not seen a dramatic increase. This is hardly surprising. Purchasing a system does not in itself mean a change in management practices.
Digitisation without integration often merely provides faster access to conflicting information.
Agent-based AI isn’t just another chatbot
This is where agent-based AI comes in. The term sounds like something concocted by the marketing department, but the concept behind it makes sense.
Agent-based AI can operate alongside existing systems. It does not need to replace BIM, ERP, scheduling or documentation platforms. Its role is to link information, identify interdependencies and present the team with the possible consequences of their decisions.
Let’s imagine a change to the foundations whilst a bridge is being built. Today, news of this triggers a flurry of phone calls, emails, meetings and analyses. It is necessary to check the design, the schedule, the availability of materials, the cost of machinery, the impact on safety, contracts with subcontractors and potential claims.
Agent-based AI can analyse these areas in parallel and suggest several scenarios. It will not make a decision on behalf of the management or the project director. However, it can ensure that a decision is not made two weeks after its effects have already spread throughout the entire project.
This is where the real value lies. Not in producing more attractive reports, but in reducing the time between a problem arising and the organisation’s response.
The most costly risks have been evident for a long time. Nobody looks at them as a whole
Delays rarely occur out of the blue. They are usually evident in advance in data relating to deliveries, work progress, design changes and resource availability. The problem is that each piece of information goes to a different team.
The same applies to cost control. An increase in expenditure does not begin on the day it appears in the financial report. It begins earlier, when a project is amended, a delivery is postponed, and the team spends several days waiting for a decision.
AI can help piece these signals together. It may also reveal something far less comforting: the organisation had all the information, but was unable to make use of it.
For some boards, this will be a greater challenge than the actual implementation of the technology.
Without governance, the result will simply be more costly chaos
The biggest mistake would be to treat agent-based AI as just another IT project. We buy a licence, set up a pilot scheme, put together a presentation for the board and announce the transformation.
This is not a technology that can be implemented in this way. AI is, rather, a new way of operating – painstakingly precise, representing a completely different level of transformation, one that tolerates no gaps, inaccuracies or chaos.
AI will not fix out-of-date schedules, inconsistent data or contracts that block the flow of information. Nor will it clarify on its own who is responsible for an incorrect recommendation and who has the authority to approve it.
Clear rules on accountability, human oversight, control over access to data, the ability to trace the decision-making process, and safeguards in the event of a failure or cyber-attack are required.
In the case of public and critical infrastructure, this is not an optional extra to the project. It is a prerequisite for its commissioning.
First a problem, then a major transformation
A sensible implementation should not start by asking where AI can be applied. It is better to ask where the organisation regularly loses money because decisions are made too late.
This could involve managing project changes, identifying the risk of delays, or coordinating deliveries. We select a single process, connect the key data sources and measure the outcome. Decision time. The number of risks detected in advance. The value of corrections avoided. The accuracy of forecasts.
Only then is it worth scaling up the solution.
AI is a brilliant way to optimise processes, but it is also extremely demanding. The companies that will gain the upper hand are those that can integrate data, identify those responsible and change the way decisions are made. The rest will end up with a very modern system for watching their project go over budget.
