Heimstaden says AI cut property-image costs by more than 99% in new OpenAI case study

Producing a virtually staged rental image now costs Heimstaden less than €0.10 in the Netherlands, down from about €50, the company says. A customer case study OpenAI published November 3, 2026 details that reduction alongside faster environmental assessments and employee adoption—company-reported results, not independently verified findings.

From employee access to rental listings

The European residential property operator began accelerating its AI effort in June 2025. After evaluating several platforms, its leadership chose OpenAI with two goals: give employees useful tools and rethink core business processes. Increasingly self-operating buildings were the ambition, rather than an already delivered result.

Heimstaden introduced ChatGPT Enterprise in September 2025. Within two to three months, half its employees were using it, according to the case study. A central AI transformation office maintains a library of use cases, allowing colleagues to learn from work already happening elsewhere in the business.

Chief Digital Officer Nikolaj Brammer describes three parts to that effort: training and tool access; a central transformation team and local AI champions; and property systems and data models prepared for AI. The letting process brings those parts together, from setting rental prices to preparing advertisements and reviewing market feedback.

AI tools now support pricing recommendations, image styling and apartment descriptions. An agent checks live advertisements against Heimstaden’s quality standards, while AI-generated market feedback helps letting teams adjust. Employees still review and approve decisions. In the Netherlands, Heimstaden estimates the cheaper image production offers €130,000 in annual savings potential—not a confirmed annual saving already realized.

An analyst builds a repeatable assessment

Environmental data was spread across public sources, formats and datasets. Julia Erlandsen, a sustainability analyst with no software development background, used Codex to help assess the portfolio. The tool helped her find, clean and standardize data, check its quality and set up repeatable steps.

Erlandsen retained control of the methods and checks. Heimstaden says the project covered 20,112 property assets and produced more than six million rows of useful data. The company estimates the delivery requirement fell from five people working for a year to two people working for two months.

Heimstaden says the project also spared it from hiring outside consultants. The team can continue to develop its own method for climate-risk assessments, reporting and investor decisions. The work can therefore support future needs, not just one assessment.

The next move reaches into business systems

Use also extends to property caretakers. In Norway, one caretaker used ChatGPT to assess a neglected apartment and draft a prioritized refurbishment plan for review and refinement. Heimstaden says staff must use approved tools under data-security rules and review AI-generated content before sending it.

Heimstaden is rolling out a Procurement Intelligence Platform that pairs automated workflows with OpenAI models accessed through APIs. It is designed to operate across Outlook, SharePoint and Teams to help analyze contracts, prepare for negotiations and manage contracts through their lifecycle. Employees remain responsible for commercial decisions and approvals.

Heimstaden is also co-developing Amplio Estate, a property-management system intended to connect AI to core business tasks through MCP. The goal is for employees to use tools such as ChatGPT to find property information and act in business systems, with human oversight. These capabilities remain a development goal.