Heimstaden Cuts Image Costs by Over 99% with ChatGPT Enterprise and Codex

Hi everyone, it's Shii-chan!

Today OpenAI shared a customer story from a European real estate company. What happens when AI joins the work of managing buildings? Let's take a look!

What was announced?

In OpenAI News, OpenAI featured Heimstaden, a European residential real estate operator. Heimstaden chose OpenAI in June 2025 after evaluating several AI platforms, with the goals of empowering employees and rethinking core business processes.

The long-term aim is increasingly self-operating buildings, where AI helps anticipate needs, coordinate work, and support decisions across the property lifecycle.

Why it matters

This is a customer story rather than a new feature, but the numbers are concrete. An analyst without a software background built data pipelines with Codex, and on-site staff use ChatGPT day to day. AI is no longer only for developers.

What changes

ChatGPT Enterprise

Heimstaden launched ChatGPT Enterprise in September 2025, and within two to three months half of its employees were using it. A central AI transformation office keeps a library of use cases so teams can learn from each other.

In the letting process, AI is used like this:

Pricing: tools built in ChatGPT Sites
Advertising: image styling and apartment descriptions
Reviewing live listings against Heimstaden's quality standards
More dynamic pricing based on market feedback

Employees still review and approve decisions.

In Norway, a caretaker who was initially skeptical used ChatGPT to assess a neglected apartment and draft a prioritized refurbishment plan, which they could then review and refine.

Codex

Julia Erlandsen, an environmental sustainability analyst with no software development background, used Codex to assess environmental risks across the whole portfolio. Codex helped with discovering, cleaning, and standardizing scattered data, quality checking, and building repeatable processing pipelines, while she kept control of the methodology and validation.

Dive Deep

Key numbers

Image production cost (Netherlands): from about 50 euros to under 0.10 euros per virtually staged image, a reduction of more than 99%
Annual savings potential: 130,000 euros in the Netherlands alone (Heimstaden's estimate)
Assets assessed with Codex: 20,112, with more than six million rows of useful data
Delivery effort: from five people for a year to two people for two months (Heimstaden's estimate), and no external consultants were needed
ChatGPT Enterprise adoption: 50% of employees within two to three months of rollout

The capability now supports climate-risk assessments, reporting requirements, and investor decisions.

Next stage: from answering to acting

Procurement Intelligence Platform (being rolled out): combines workflow automation with OpenAI models accessed through APIs. It works across Outlook, SharePoint, and Teams for contract analysis, negotiation preparation, and contract lifecycle management. Employees keep responsibility for commercial decisions
Amplio Estate: a property management system Heimstaden is co-developing. It is designed to support AI workflows through MCP, so employees can use tools like ChatGPT to retrieve property information and take action in core systems, with human oversight

Responsible adoption

Employees use approved tools under data security rules and must review AI-generated content before sending it. Andreas Kalmar Lindquist, Manager of the AI First Team, says the goal is to scale what works while keeping data secure, responsibilities clear, and people in control.

Leadership lessons

Focus on people: adoption and imagination are the bottlenecks, not technology
Find champions: the biggest returns come from colleagues driving change locally
Fix the process first: AI will not repair a broken process, but it can greatly scale a good one

Wrap-up

Heimstaden launched ChatGPT Enterprise in September 2025, and half of employees were using it within two to three months
In lettings, AI supports pricing, advertising, and listing review, and image costs fell by more than 99%
With Codex, an analyst without a development background assessed 20,112 assets for environmental risk
Next up: AI that acts, including an MCP-ready property management system
People stay in charge of decisions and responsibility

This story is for anyone driving AI adoption inside a company, and for anyone curious about helping non-engineers improve their own work!