Planning AI Watch is a weekly roundup of what public agencies around the world are doing about AI readiness and governance, selected for planners.

This week

Two threads ran through this week. In England, AI-assisted planning is moving from trial toward general availability, with PlanAssist expanding to 15 councils and a Budget commitment to offer it to every council by the end of 2027, while a Dutch government audit shows how easily algorithm governance stays on paper, especially for purchased tools. Closer to the permit counter, Everett and Pflugerville show two practical pieces of the same puzzle: narrow tools that help applicants submit complete applications, and local rules that keep people accountable for significant decisions. A UK study of bot-written consultation responses is a reminder that public comment processes need their own safeguards. Meanwhile, data center siting stayed on planning agendas on three continents, with councils weighing moratoriums against case-by-case review and asking who bears the water, power and land costs.

Worth your time

UK expands PlanAssist trials to 15 councils as new delegation rules take effect

Policy | National | England, UK (MHCLG and HM Treasury, GOV.UK) Twelve more councils, including Lambeth, Newcastle, Nottingham, Bristol and Westminster, join Camden, Barnet and Dorset in a 12-month trial of the government’s PlanAssist AI tool. The government says early testing cut time on key planning tasks by 58%, that planning officers remain responsible for reviewing applications and making decisions, and that PlanAssist will not be mandated. From October 31, straightforward applications are decided by planning officers under new delegation rules. Separately, the 2026 Budget commits, subject to further testing, to making PlanAssist available to all councils in England by the end of 2027, with the aim of halving routine decision times, and to mandating publication of key planning documents as open data.

Why it matters: A national government is now putting an AI planning tool on a path to every local authority, so councils elsewhere will want to watch what trial evidence is published and how officer accountability holds up in practice.

Dutch audit finds the national algorithm framework is often a paper exercise, and purchased algorithms a blind spot

Report | National | Netherlands (Gemeente.nu) Two studies by the Auditdienst Rijk, the central government audit service, found the national Algoritmekader was largely unknown beforehand in six of eight government organizations reviewed. The impact assessment field was empty in about two-thirds of algorithm register entries, and five of ten organizations had no process for registering algorithms they buy. The interior ministry says it will act on the recommendations. The audits covered national bodies, but the framework is also meant for municipalities and other public bodies.

Why it matters: Registers and frameworks only build public trust if they are used, and procurement is where many agencies lose sight of the algorithms they depend on.

Pflugerville, Texas, adopts an AI and surveillance technology policy with a resident review commission

Policy | City | Pflugerville, TX (Community Impact) The City Council unanimously approved its SAFE Technology policy on October 27. AI and automated systems may assist staff but cannot replace human decision-makers in significant matters such as code enforcement, permits and access to city services. Outside parties may not sell city data or use it for unrelated purposes, including training general-purpose AI models, without council approval. A resident commission appointed by the council will review proposed technology, and the city will publish an annual report on high-risk technology and its safeguards.

Why it matters: It is a compact local model that names permits directly and pairs human decision rules with data limits and public reporting.

Everett, Washington, pilots AI intake help for ADU permit applications

News | City | Everett, WA (The Everett Herald) Everett is piloting a tool from a Seattle startup that shows accessory dwelling unit applicants which documents their project needs, checks that required information is included, and translates city codes and checklists into eight languages. City data shows about 95% of permit applications from January 2023 to June 2026 needed at least one correction. Public launch is expected in spring 2027. The company says its AI can only make suggestions, and people make all decisions.

Why it matters: Incomplete submissions are a familiar source of permit delay, and a narrow intake pilot with clear human decision-making is a measurable place to test whether AI helps.

Study finds bots wrote about three quarters of responses to a UK government consultation

Research | National | United Kingdom (PublicTechnology) A study in the journal Tobacco Control found that almost 91,000 of close to 119,000 responses to a 2023 Department of Health and Social Care consultation on smoking and vaping were identified as automated, mostly opposing limits on disposable vapes. With those responses excluded, three quarters of valid respondents supported change. The researchers could not independently check the department’s detection methods, and their request for respondent metadata was ruled ineligible under freedom of information rules.

Why it matters: Planning agencies rely on public comment, and this is a concrete warning that consultation processes need ways to detect automated submissions and to explain how they did so.

Data center siting

Councils and national governments kept working through where data centers belong and on what terms, with several moving from by-right approval to case-by-case review.

  • San Antonio: City Council voted 9-2 to end by-right data center development; each new data center now needs a specific use authorization from council, is limited to commercial or industrial zones, and is reviewed for distance from neighbors, noise, water and energy use and lighting (Texas Public Radio)
  • San Diego: The council unanimously adopted a 45-day moratorium that could be extended for nearly two years, and gave staff 180 days to bring recommendations, guided by a stakeholder working group, ahead of a land development code amendment (KPBS)
  • New Hampshire: Lawmakers requested 2027 bills, including one that would define data center sizes in state law so towns can regulate by scale; its sponsor also intends to propose banning nondisclosure agreements between developers and municipal officials during permitting (Route Fifty, from New Hampshire Bulletin)
  • Victoria, Australia: The Municipal Association of Victoria, representing all 79 councils, resolved to seek a moratorium on new or expanded data centre approvals until a comprehensive planning and environmental framework exists, and criticized a state fast-track that lets applicants bypass councils (ABC News)
  • Malaysia: The national government said states without enough water supply cannot host data centres even where land is available, and ruled out a blanket moratorium in favor of judging each application on water, electricity and investor finances (The Edge Malaysia)
  • England: The Budget introduced a Data Centre Covenant tying government grid support to commitments on nature, water use, annual energy and water reporting and local benefits, and opened consultation on a draft national policy statement for data centres (Pinsent Masons Out-Law). Separately, a freedom of information investigation found officials met 137 businesses and no environmental groups while shaping AI growth zone policy (Resultsense)

Also noted

Planning AI Watch is published by SchillerTech Group, an independent advisory that helps planning departments decide whether and when to adopt AI, on their own terms. Seen something we should include? Write to tom@schillertech.com.

Discover more from SchillerTech

Subscribe to get the latest posts sent to your email.