AI for better and more efficient decisions in real estate asset management

Deutsche Telekom AG has a considerable real estate portfolio with more than 10,000 of its own buildings and rental properties. Optimizing usage offers the potential for long-term savings of several hundred million euros.

While decisions are currently mainly made manually by real estate experts, Deutsche Telekom's innovation department is now developing a web-based tool that offers AI-based decision support for asset strategy planning.

The main problem with portfolio strategy development is the high complexity of many different boundary conditions and factors to consider, including: rental agreement periods, limited capacities, critical technical infrastructure, possible relocations, allocation of tenants and companies, works councils, and desk-sharing quotas.

The challenge is that the number of possibilities grows exponentially with the number of buildings considered. In addition, there are different tenants for whom it is necessary to plan suitable compensatory areas and relocations. It is often difficult to find ideal solutions, even for experts, and alternative scenarios for comparison are time-consuming to create.

Building on initial algorithms designed by data scientists at Deutsche Telekom, the data science experts at codecentric AG, together with the project team at Deutsche Telekom and in cooperation with other departments at the client, developed new optimization algorithms for portfolio strategy planning and supported their integration into the microservice architecture. Continuous coordination with end-users allowed their feedback to be meaningfully integrated into the optimization logic, avoiding possible skepticism about this new technology.

Further advisory services provided by codecentric AG in the data science area covered topics to ensure data quality and data consolidation of the various sources.

The developed tool (AI-REM), whose license rights are held by Telekom, supports real estate experts in validating existing lease termination projects and relocation strategies. Thanks to the support of the data scientists at codecentric AG, it is now for the first time possible to offer decision support tools that are theoretically optimized for a limited number of assumptions and constraints.

The AI-REM tool received positive feedback within Telekom and at public presentations, so it will be further developed for general portfolio management and can be seen as a successful example of how the development of intelligent digital assistants can implement artificial intelligence in a company effectively.