AI in Real Estate in Sweden: What Property Companies Can Use AI For and Where to Start

What can Swedish property companies use AI for?

A property company runs on three kinds of information: technical data from buildings, conversations with tenants, and documents such as leases, drawings, invoices and energy performance certificates. AI helps wherever one of these is large, repetitive and handled today by people reading, sorting or copying.

The sector is already experimenting. A 2026 survey by Fastighetsägarna, Fastigo and Sveriges Allmännytta found that close to six in ten organisations use AI tools, mainly for text work, financial administration, decision support and automated property management. Only 17 percent had guidelines for AI, and 60 percent said they lacked internal knowledge. The gap is not tools but choosing problems and handling data.

Source: Alice Labs assessment; regulatory context from Boverket, IMY and the EU AI Act

What each use case involves

In short

Energy AI works on data Swedish rules already require in larger non-residential buildings. Tenant service and lease analysis use language models and are the quickest to test. Rent, design and ESG work depend on how clean your own data is.

Energy and maintenance. Since 1 January 2025, Swedish non-residential buildings with heating, ventilation or cooling systems above 290 kW must have building automation and control systems, which Boverket says should monitor, record, analyse and adjust energy use. Their trend logs let a model run forecast-based heating control, find systems running outside opening hours and flag equipment close to failure. Start where the automation already exists.

Tenant service. A language model can classify, prioritise and route free-text fault reports and answer common questions. Keep a person in the loop for safety, water damage and access to homes, and tell tenants when they are chatting with AI.

Leases and documents. AI can extract index clauses, extension terms and cost allocation into a register and compare drafts against your standard terms. A lawyer checks the output, so risk is low.

Rents and markets. Most Swedish residential rents are negotiated collectively under the utility value system, so AI pricing matters mainly for commercial premises. For housing, use it for analysis ahead of negotiations.

Design and ESG. New buildings already need a climate declaration and the EPBD adds life-cycle climate requirements, so assessing design options early matters more. See our guide to AI in construction for the builder's side. ESG reporting is a good early project: AI extracts and reconciles figures, a person signs off.

How to choose the first AI project

In short

Score each candidate on business value and on feasibility: data availability, integration effort and regulatory risk. Start in the high-value, high-feasibility corner, which for most Swedish property companies means energy data, fault reports or leases.

A first project should create measurable value and teach the organisation how to run the next one. Score each idea from one to five on value (energy cost, staff time, revenue, risk) and feasibility (data in hand, integration effort, legal exposure), then pick one from the top-right quadrant. Three questions settle most debates: do we have twelve months of usable data today, can a person check every output during the pilot, and could we explain the system to a tenant, to IMY and to our board? If any answer is no, it is not the first project.

Source: Alice Labs method

Data and GDPR: handling tenant data correctly

In short

Fault reports, tenant messages, lease registers and apartment-level energy data are personal data, so GDPR applies from the first prototype. Follow IMY's guidance on GDPR and AI: purpose, legal basis, minimisation, processor agreements and a DPIA check.

Fault reports describe life inside a home, apartment-level energy data shows when someone is there, and a lease register links it all to a person. IMY, the Swedish Authority for Privacy Protection, publishes guidance on GDPR and AI that covers development and testing, not only production. In practice:

• Define the purpose of each AI use and check it is compatible with why the data was collected.
• Document the legal basis, often legitimate interest or performance of the lease.
• Minimise: pseudonymise fault reports and aggregate energy data to building level where possible.
• Sign data processing agreements with AI vendors and check where data is stored and processed.
• Assess whether a data protection impact assessment is needed, especially for profiling or systematic monitoring.
• Respect GDPR limits on decisions based solely on automated processing.

None of this blocks the recommended first projects. It does rule out pasting tenant data into consumer AI tools.

The EU AI Act for property companies

In short

Energy optimisation, document analysis and fault triage are normally not high-risk. Credit scoring of natural persons is high-risk under Annex III, and the Act cites access to housing. Annex III rules apply from 2 December 2027. AI literacy and chatbot transparency duties already apply.

Tenant screening can be high-risk. Annex III lists AI used to evaluate the creditworthiness or credit score of natural persons as high-risk, except for fraud detection. Recital 58 explains that such systems determine access to essential services such as housing. Whether a given screening tool is covered depends on what it does, so treat it as potentially high-risk until a lawyer concludes otherwise. After the AI Omnibus, in force since 27 July 2026, Annex III rules apply from 2 December 2027. Deployers then need human oversight, logging, use according to the provider's instructions and, in some cases, a fundamental rights impact assessment. Swedish discrimination law applies regardless.

Transparency. A chatbot on the tenant portal must make clear that the tenant is talking to an AI system.

AI literacy. Article 4 has applied since 2 February 2025 and requires measures so that staff using AI have sufficient AI literacy. Given how few property companies have AI guidelines, most are behind here.

Swedish context: Boverket, Energimyndigheten, energy performance certificates and the EPBD

In short

The recast EPBD (Directive 2024/1275) aims for a zero-emission building stock by 2050. Boverket leads Swedish implementation with support from Energimyndigheten. Energy class A0 arrived on 25 May 2026, law changes on 1 July 2026 and further rules are set for 1 October 2026.

Regulation is the strongest reason to start with energy AI. The recast Energy Performance of Buildings Directive entered into force on 28 May 2024 and was due to be transposed by 29 May 2026. New public buildings must be zero-emission from 2028 and all new buildings from 2030. Member states must set minimum energy performance standards for existing non-residential buildings and trajectories for residential ones, and new buildings must be ready for solar installations.

Boverket is the responsible authority and Energimyndigheten contributes energy expertise. Boverket's timeline says Swedish implementation is somewhat delayed and comes in stages: energy class A0 and a temporary zero-emission building definition from 25 May 2026, changes to the Planning and Building Act and the Energy Performance Certificates Act from 1 July 2026, and revised Boverket rules on energy management and certificates from 1 October 2026.

For owners, certificates and meter data become strategic inputs to renovation and financing decisions, and automation data required for compliance can also drive optimisation.

A 90-day plan for the first AI project

In short

Days 1 to 30: inventory data, pick one use case and set a baseline. Days 31 to 60: run a narrow pilot with people checking every output. Days 61 to 90: measure against the baseline, decide whether to scale and document governance.

Days 1 to 30: choose and prepare. Score candidates in the matrix and inventory data for the top two: where it lives, who owns it, whether it contains personal data. Pick one and set the baseline, such as energy per square metre, time to route a fault report or hours per lease review. Agree legal basis and processor agreements, and publish a short AI policy.

Days 31 to 60: build a narrow pilot. Limit scope to a few buildings, one fault-report queue or one lease type, using existing systems where possible. The people who do the work today check every output and log disagreements.

Days 61 to 90: measure and decide. Compare against the baseline, record the data problems you found, then scale, adjust or stop. Scaling needs an owner and a review cycle.

How Alice Labs works with property companies

In short

Alice Labs helps property companies choose the right first AI project, check data and GDPR readiness, build a measured pilot and scale what works.

We start with a short discovery: interviews with asset management, property management, technical operations and finance, and a review of your existing data. The output is a prioritised use-case list and a recommended first pilot with what it takes to run it lawfully. We then build the pilot with your team, measure it against an agreed baseline and hand over GDPR and AI Act documentation, including AI literacy measures. We are vendor-neutral.

Read more on our Swedish page on AI for the real estate sector, or see our approach to AI strategy and AI automation.