Heimstaden uses ChatGPT and Codex to take over parts of property management

Heimstaden, a European residential real estate operator, is using OpenAI's ChatGPT Enterprise and Codex across letting, sustainability analysis and procurement. The details come from a customer case study that OpenAI published on 3 November 2026. All figures in it come from OpenAI and Heimstaden and have not been independently verified.

The numbers matter because property management is mostly coordination: pricing a flat, photographing it, writing the advert, checking listings, assessing risk, reading contracts. These are the tasks Heimstaden is handing to AI. Employees stay in place to review and approve the results.

Letting homes with less handwork

According to OpenAI, Heimstaden's leadership decided in June 2025 to speed up its use of AI. It looked at several platforms and chose OpenAI. Co-CEO Helge Krogsbøl calls AI "one of the biggest opportunities to accelerate that transformation" of how the buildings are run.

The company started with the letting process, the work of getting an empty home back on the market. AI now supports four steps:

• Pricing: recommendations from tools built in ChatGPT Sites.
• Advertising: AI styles the images and writes the apartment descriptions.
• Listing review: an agent checks every live advert against Heimstaden's quality standards.
• Feedback: AI-generated market feedback helps letting teams correct prices, which the company says makes pricing more dynamic.

In the Netherlands, the cost of one virtually staged property image fell from about €50 to under €0.10, according to the case study. Heimstaden puts the annual savings potential in that market alone at €130,000.

Chief Digital Officer Nikolaj Brammer describes three parts to the strategy: training and tools for employees, a central transformation office with a network of "AI champions", and preparing the company's property management systems and data models for AI. "It's not about automating one task, but improving how the whole process works," he says of letting.

A sustainability analyst writing data pipelines

The biggest single result in the case study comes from outside software engineering. Heimstaden needed to assess environmental risk across its portfolio. The data was there, but scattered across public sources, formats and datasets.

Julia Erlandsen, an environmental sustainability analyst with no background in software development, used Codex to find, clean and standardise the data, check its quality and build repeatable processing pipelines. OpenAI says she kept control of the methodology and the validation.

The project assessed 20,112 assets and produced more than six million rows of data. Heimstaden estimates the work would otherwise have taken five people a year. It took two people two months. The company also did not need the external consultants it would otherwise have hired. The output now feeds climate-risk assessments, reporting and investor decisions. "We now own the methodology and can continue developing it ourselves," Erlandsen says.

Adoption, rules and the next systems

ChatGPT Enterprise went live at Heimstaden in September 2025. Within two to three months, half of employees were using it, according to the case study. The central office keeps a library of use cases. Staff must use approved tools under data security rules, and they are responsible for checking AI-generated content before sending it and for approving decisions.

The tools also reach caretakers, some of whom were sceptical at first. In Norway, one caretaker used ChatGPT to assess a neglected apartment and draft a prioritised refurbishment plan, which they then reviewed and refined.

Two systems are meant to move AI from answering questions to taking action. The first is a Procurement Intelligence Platform, now being rolled out. It combines workflow automation with OpenAI models accessed through APIs, runs across Outlook, SharePoint and Teams, and supports contract analysis, negotiation preparation and contract lifecycle management. Employees keep responsibility for commercial decisions and approvals.

The second is Amplio Estate, a property management system Heimstaden is co-developing. It is designed to support AI workflows through MCP, the Model Context Protocol, so that employees can use tools such as ChatGPT to retrieve property information and act in core business systems "with appropriate human oversight." The company's stated long-term aim is "increasingly self-operating buildings."

The case study does not say how many people Heimstaden employs, how many homes it manages in total, or whether any roles have changed or been cut as a result. It reports savings potential and time estimates, not audited results.

What it means for work

In letting, the production work around a listing, such as staging photos, writing copy and checking adverts against standards, is moving to AI. Letting staff now mostly review and approve. The clearest shift is in analysis: a domain expert who cannot code built pipelines that Heimstaden says would have needed five people for a year, and the company kept the work in-house instead of paying consultants. That points to less demand for outside data and consulting help, and more weight on staff who can define a method and check what the machine produces. Caretakers and procurement teams are next. AI will draft plans and prepare negotiations, and people keep the decisions. What remains open is whether the "self-operating buildings" plan, once AI can act inside core systems, keeps the same number of people in those review roles, and how well that oversight holds up as volumes grow.