PT
Imagem editorial para Antes da primeira pedra
Published in Issue 6 — Before the first stone

Using AI for planning permission is not the same as letting AI decide

The pilot projects announced for planning permission may reduce repetitive tasks. Technical and administrative responsibility, however, cannot be automated.

Artificial intelligence can help check documents, cross-reference rules and organise applications. Any time saving will only be credible if there are consistent data, transparent criteria and human decision-making.

Composição editorial abstrata sobre cadastro, licenciamento, reabilitação e construção habitacionalIssue 6 — Before the first stoneRead the complete publication

Category: Construction and Infrastructure

The pilot projects announced for planning permission may reduce repetitive tasks. Technical and administrative responsibility, however, cannot be automated.

Portugal has launched municipal pilot projects to apply artificial intelligence to planning permission. The stated goal is to bring decision times closer to 50 days. This ambition addresses a problem familiar to developers, designers and municipalities: lengthy procedures, scattered documents, repeated requests for information and rules that are not always interpreted consistently.

Technology can be useful before a decision is made. A system can confirm whether an application contains the mandatory documents, identify inconsistencies between forms, compare areas and parameters, or direct the application to the relevant departments. It can also flag applicable regulations and produce a checklist for the municipal officer. These tasks consume time and are sufficiently structured to benefit from automation.

The limit is reached when checking is confused with deciding. A project is not simply a collection of fields. It involves the urban context, heritage, accessibility, public utilities, safety and effects on third parties. Municipal regulations may contain concepts that require interpretation. Drawings and design statements may be formally consistent yet technically insufficient. The final decision must therefore remain identifiable, reasoned and signed by someone with legal authority.

Data quality will be decisive. If municipalities use outdated databases, incompatible mapping or poorly structured rules, artificial intelligence will merely accelerate errors. It must also be possible to know which version of each regulation was used, what checks were carried out and why an alert was generated. Without an auditable record, applicants cannot understand or challenge the course of the assessment.

Responsible use should begin with low-risk tasks and publicly measured results. Average processing time, the number of requests for clarification, return rates and discrepancies between automated assessment and human decisions should all be compared. A shorter timescale is not enough if it leads to more subsequent corrections, disputes during construction or different treatment of equivalent cases.

Digital planning procedures can make the process predictable. To achieve this, technology must support technical teams rather than replace planning criteria or conceal responsibility.

Key points

1. AI can check documentation, cross-reference parameters and organise applications before technical assessment. 2. Decisions must remain human, reasoned, auditable and subject to the legal powers of each authority. 3. Success should be measured by timescale, quality, the number of returned applications and consistency, not merely speed.

Note

Automating the initial checks can free up technical time for the aspects that genuinely require interpretation, coordination and responsibility.

Closing

The best application of artificial intelligence to planning will probably be the least spectacular: fewer lost documents, fewer repeated requests and greater clarity about the status of each application. If technology makes the route to a decision visible, 50 days can become a service target. If it merely produces faster answers without explanation, it will be another layer of opacity.

Sources

idealista/news / Lusa, 21 September 2026. AI in planning applications? Government proceeds with pilot projects.

Read the complete publication

In this issue

Issue 6Housing has moved beyond reach three times faster than wagesThe gap between housing prices and incomes has widened again. The problem is no longer limited to the price per square metre: it involves the deposit, mortgage payments, location and time.© MATTERS / FotocomposiçãoIssue 6There can be no land policy without a land registerInventorying assets and locating properties does not automatically create available land. It does, however, provide the basis for deciding what can be transferred, rehabilitated, protected or developed.© MATTERS / FotocomposiçãoIssue 6A factory need not be erased to house studentsResidência Universitária Confiança almost doubles the Universidade do Minho’s accommodation capacity in Braga and shows how an industrial building can house a contemporary programme without losing its urban memory.© MATTERS / Fotocomposição

Related

Issue 5More capacity, fewer buildingsConstruction employs more people and receives more financing, but these signs of activity have yet to translate into equivalent growth in supply. In the second quarter of 2026, both building permits and completions declined.© MATTERS / Composição gráficaIssue 5Refurbishment sustains work but loses its project pipelineUrban refurbishment activity increased by only 0.8% in July. In the first half of 2026, 2,960 refurbishment projects were licensed, 12.5% fewer than in the corresponding period.© MATTERS / Ama Grupo Imobiliário®Issue 4The housing shortage is also an execution problemPortugal is building below the level the sector estimates is needed. Between land and a completed home lie decisions, permits, financing, productive capacity and time.Imagem criada para MATTERS.