Building robust foundations for effective AI in real estate

Being specific, AI systems rely on data to learn, adapt, and make predictions. To trust these predictions, there are quality standards that the data must meet. The first is accuracy – if an AI model is fed incorrect data, it will produce incorrect outcomes. Next, data needs to be representative – up to date, specific to the problem you’re trying to solve – but also diverse – detailed and broad enough to capture the nuances of the problem. The larger the volume of training data, the higher the likelihood of AI outcomes being accurate, reliable, and free from bias. Simultaneously, with a large knowledge base at their disposal, AI tools can expand the opportunity to combine and manipulate big data, creating superior outcomes.

Effective data management practices, including data storage, cleaning, normalisation, and integration, are also essential to building a trustworthy AI system.

At CBRE, we have implemented AI across a range of business functions and teams to drive not only efficiencies but new revenue-generating opportunities for us and our clients. Through this process, we have learnt that while data foundations are crucial to AI implementation, they are one of four key foundations, alongside:

• Infrastructure: AI requires robust technology infrastructure, including powerful hardware and scalable cloud solutions. The right infrastructure in place means that AI models can be trained efficiently and deployed at scale, with the necessary support for data handling, where data can be stored, processed, and accessed efficiently. This is vital for AI operations that typically involve large volumes of data, and so investing in the right infrastructure is crucial to supporting the demands of AI projects.
• Integration: Successful AI implementation requires seamless integration with existing systems and workflows across an organisation. This ensures that AI solutions complement and enhance current processes rather than disrupt them, being brought into a workflow when they are needed so that they are not separate from the experience. This is true from a technology perspective, where AI systems can leverage data from various sources, and from a human perspective, as effective integration aids in incorporating AI into a colleague’s day-to-day role. This thinking needs to go beyond individual tools and take a holistic view. There’s no ‘one size fits all’ when it comes to businesses and teams using AI, and so we must ensure that we’re integrating AI in a way that is helpful and useful to the humans we’re trying to support.
• Scalability and flexibility: As AI systems continue to evolve, the data foundations they are built on must be able to adapt and scale accordingly. Similarly, as business needs evolve, AI solutions must also be able to scale and adapt in response. Echoing our earlier point on taking a holistic and human-centred approach to AI integration, each business and team must ensure they are focussing on where AI will be most impactful for them – be it customer service chat bots or summarising documents – and flexing it to their needs.

However, even with these four key foundations, success is not guaranteed. People, process, and prompting also play crucial roles:

1. People: Our employees' roles are evolving to accommodate the increasing prevalence of AI in our operations. We are guiding them through this transition through four strategic areas of focus: education, experimentation, implementation, and ethics. Focussing on our first two pillars, upskilling, and building confidence in AI are key to this evolution. We have established upskilling and change programmes to help our colleagues embrace AI and adopt a more data-centric approach, such as an AI Champions Community, a range of training courses and resources, and regular communications to offer inspiration and guidance.
2. Process: Data management, governance, and ethics are paramount. Data needs to be handled responsibly and ethically, complying with regulations, and building trust with our clients. Publishing legal documentation isn’t enough, and ethical guidance should be provided through a range of digestible and relevant resources and training. Of course, the process is also crucial in project conception. It is essential to have clear objectives and use cases, and a unified focus on tracking business outcomes and value. By identifying what we want to achieve with AI and how we will monitor progress, we can ensure AI implementations are aligned with and help to achieve broader strategic goals.
3. Prompting: This refers to the interaction between users and AI tools, and the instructions that our colleagues give to their AI tools. The best data in the world would be useless if not used effectively, so we have upskilled and supported our colleagues to write more effective prompts, and we continually improve the interface and interaction between our users and our AI tools.

Implementing AI requires more than a one step process, or a one size fits all approach. Without quality data, outcomes derived from AI could be meaningless. But conversations around good data also emphasise the role that humans need to play during the transition. Organisations adopting AI need strategies underpinned by strong communication, formalised processes, and human validation to upskill their business and ensure they achieve effective outcomes.