More and better railway, in half the time

The target is the industry's own

The industry's own. Omtag Svensk Järnväg is an initiative started by SJ and run through Tågföretagen, gathering train operators, contractors and public transport bodies around one question: how Sweden gets more railway for the money. In the 2025 report Skenande kostnader, economists Fredrik Bergström and Tore Englén quantified that ambition for Innovationsföretagen in a single line:

Mer och bättre järnväg på halva tiden till halva kostnaden.

Not Trafikverket's target, and not the government's. The distinction matters: a sector that sets its own bar is volunteering to be measured against it. And the bar is less fantastic than it sounds. Swedish rail investment prices have grown 4.0 percent per year since 1990 while consumer prices grew 2.1 percent, according to the same report. Today's cost level is the result of 34 years of drift, not a law of nature. Halving it is a return journey, not science fiction.

The slow years are the planning years, not construction

Not the construction years. Riksrevisionen, Sweden's national audit office, examined 285 projects in the national transport plans and found no systematic cost deviations in the build phase: completed projects landed, on average, within a few percent of their build-start estimates (RiR 2021:22). The planning years are another story. Projects that sat in the plan between 2014 and 2018 grew 39 percent in estimated cost over those four years. The follow-up audit (RiR 2023:25) found that projects still in the planning phase grew 51 percent between the 2018 and 2022 plans, against 16 percent for projects already under construction. The total growth in those four years, 112 billion kronor, roughly matched the entire new appropriation for the same period, 111 billion.

Jonas Eliasson's 2025 study of Swedish national-plan projects sorts the growth by phase: 24 percent in early planning, 27 percent in late planning, then 4 percent at the build decision and 3 percent during construction. Across 78 Dutch projects, Cantarelli and colleagues measured 5 percentage points of additional overrun per extra pre-construction year. The halving question, in other words, is not about pouring concrete faster. It is about the planning years.

What fills those years is document work: investigation rounds, review cycles, redesign, permit documentation. Produced and checked by hand, at human speed.

Technology can reach the repetitive document half

The repetitive document work, and that is most of it. A railway plan is not written from a blank page; it follows the same governing documents as the last one, and the large majority of Trafikverket projects are variations on that settled theme: the same approach and the same governing documents, project after project.

Standardized, repetitive document work is precisely what AI systems now do well. Be precise about what that means: not less scrutiny, and not shorter deliberation. The consultation rounds, the participation and the political decision points all stay. What shrinks is the production time between them. A review round that ties up engineers for weeks can run in days; the documentation for a design alternative that used to cost three weeks costs an afternoon. Cheaper iterations mean more alternatives actually get evaluated before decisions lock in — and lock-in of premature decisions is, in Eliasson's analysis, the mechanism behind the planning-phase growth itself. His remedy is to keep more projects and designs in genuine competition, longer. That requires analysis capacity, and machine-speed document work is where the capacity comes from.

The decision system needs policy, not tools

The decision system. Norway shows both halves clearly. Under its finance ministry's external quality-assurance regime, every large project is reviewed before parliament commits the money, and Welde and Odeck found that Norwegian road projects came in about 2 percent under budget in the implementation phase. The same study found that estimated costs had grown about 40 percent during the planning stage, before the gate could catch them. Governance repaired the back half of the timeline. The front half is a tooling gap.

Sweden's policy half is documented in its own audits. The government rarely reconsiders a project once it has entered the national plan: "regeringen håller ofta fast vid sitt första investeringsbeslut", as Riksrevisionen put it — the government often holds on to its first investment decision. The 2023 audit also found no correlation between a project's socioeconomic return and whether it was selected for the plan. No document tool fixes that. Gates, re-tested decisions and late selection do.

So read the vision the way it deserves: not as a promise, but as a target the sector can work toward from both ends. Nobody can promise halves, and this post does not. But the half of the timeline that consists of producing and checking documents is addressable with technology that exists today, and the governance half has a working example across the border. Yesper is the AI civil engineer for construction and infrastructure. The document half of the railway timeline is precisely where it works.