From Three Months to One Week: How the City Planning Office Streamlined Detailed Development Planning with AI

What happens when three months of manual document work can be cut to one week? At the City Planning Office of Stockholm City, AI has become a practical support in the detailed development planning process.

The Planning Department at the City Planning Office of Stockholm City manages an extensive workload. Detailed development planning for large projects can generate over a thousand statements during both the consultation phase and the review phase — from referral bodies, organisations, and private citizens. Compiling, interpreting, and assessing these statements was previously a cumbersome process that tied up a significant amount of time for each case officer. In addition to the consultation report/review statement, a service memorandum is also produced, involving multiple stages of internal review.

A specific consultation report could previously take three months of full-time work to produce. For service memoranda, two weeks per case was the norm. Case officers manually read through emails, interpreted the content, condensed the text, and registered it in the system — a process that was both time-consuming and mentally exhausting.

What They Did

In mid-May, the Planning Department opened the Intric assistant to the entire department and began experimenting with two specific tasks:

1. Consultation Reports / Review Statements

The assistant was tasked with sorting, summarising, and evaluating hundreds of statements. However, it was critical to get the balance right between different types of input: all statements had to be assessed correctly in accordance with guidelines and relevance.

Through iterative development, they arrived at a solution that handled this ambiguous assessment effectively. Statements were compiled in accordance with the correct instructions and evaluations, in the correct structure.

2. Service Memoranda

For this task, the department already had a robust internal knowledge base: hundreds of service memoranda written per year in well-established formats and writing styles, along with governing documents on how the office should phrase itself. This became the key. The assistant learned the internal style from this material and was subsequently able to draft new service memoranda that the case officers' managers could not distinguish from human-written text — several even noted that the AI-generated documents were preferable to many manually written ones.

Case officers retained control by writing all positions themselves, while the assistant handled the formulation. Some wanted feedback before the final result; others preferred to make adjustments afterwards. Flexibility was important.

After just one month, the Planning Department conducted a survey:

• Almost all case officers use the assistant daily

• Almost all experience significant time savings

• Almost all are very satisfied

After the initial measurements, the estimated effects are striking:

Consultation Reports / Review Statements:

3 months → 1 week = ~12 weeks saved

Service Memoranda:

2 weeks → 1 day = ~1–2 weeks

A case officer who previously spent three months on a consultation report and review statement can now deliver the same work in a matter of days — and with higher quality. When drawing up new time plans for detailed planning projects, the department can now reduce certain stages from three months to a few weeks. For projects planned over 10–15 years, this could mean an acceleration of several months.

The Effect Spreads

An unexpected side effect: the time saved was not limited to case officers — management also benefited, as they are now freed from language review duties.

For the City Planning Office of Stockholm City as an organisation, this potentially means that detailed planning processes can be shortened and that time is freed up for higher-quality work.

How They Succeeded

Limited initial scope. They did not try to automate everything at once. They focused on two specific, text-heavy tasks where a language model had a clear and obvious benefit.

Collaboration within the organisation. Case officers and a technically driven team member experimented together. This meant that prompts and workflows could be adapted to reality rather than the other way around.

Iterative development. They saw the effects early on but then worked iteratively to ensure quality and the desired results that work in practice within the organisation.

Awareness. They recognised that the assistant's value lay in removing the administrative burden (reading, interpretation, summarising, formulation). But decision-making and responsibility must remain with the case officer.

In practice, this means: the case officer writes the position statement (the part that requires knowledge, judgement, and responsibility), and the assistant converts it into the correct language and structure. The case officer reviews every output before it is sent. This ensures that all documentation is quality-assured, backed by genuine knowledge and authority.

Sharing and Collaboration

The Planning Department is happy to share their prompts and working methods with other municipal planning administrations. Consultation reports and service memoranda are processes that many municipalities have. There is no reason for others to reinvent the wheel.

Press contact Tiger Landén | tiger.landen@intric.ai

About Stockholm

The City of Stockholm is Sweden’s capital and the country’s largest municipality, with over one million residents. The city has around 40,000 employees and is responsible for a wide range of public services, from education, social care, and urban development to culture, environment, and democratic processes. With its historic cityscape, waterfront location, and rapid development, Stockholm is both Sweden’s political center and one of the most dynamic cities in the Nordics.

About Intric

Intric develops secure, sovereign, and compliant AI infrastructure for public authorities and critical institutions. The platform enables organisations to deploy AI assistants and agents based on their own data, under their own governance, and without dependency on external cloud infrastructure. Intric is developed in Sweden and is actively used by regulated institutions in the Nordic countries and in Germany. Founded in 2021, headquartered in Stockholm. www.intric.ai