Interview: Maria Empis - Real Estate, Risk and the Generative AI Gap
AI is quietly reshaping real estate, where data, trust and human judgement now define the industry’s next competitive edge and value creation.
Why Maria Empis believes the next phase of value will depend on trust, data quality and operational discipline
At the Lisbon MBA Católica|Nova - Alumni Unite Summit, held on 27 February 2026 at Nova SBE in Carcavelos, one of the most revealing aspects of the real estate discussion was not simply what was said, but what was left unsaid. The wider event was framed around “Navigating a World of Changing Priorities”, and the Real Estate Think Tank focused on a central question: who holds the risk in a market shaped by inflation, supply constraints, regulation and prolonged uncertainty. Yet despite the strategic depth of that conversation, technology and artificial intelligence were notably absent from the room’s explicit debate.
That absence now matters. Generative AI is no longer a speculative theme for the property sector. It is becoming an operational issue, a governance issue and, increasingly, a competitive issue. The practical question is no longer whether AI will affect real estate, but where value will be captured first, which functions will change fastest, and how firms can modernise without eroding trust.
The interview identifies this gap: while market participants remain focused on affordability, planning, supply, and cost pressures, a parallel transformation is taking shape in the information systems, workflows, and customer interfaces that support the industry.
Maria Empis, Co-Head of Residential at JLL Portugal, approaches this shift from a notably pragmatic perspective. Her answers do not suggest an industry on the verge of being replaced by generative systems, nor do they indulge in technological hype. Instead, they point to something more useful: AI as an enabler of better internal efficiency, faster service, stronger data use and more consistent client outcomes. In that sense, the real opportunity is not to make property more artificial, but to make the organisation behind it more intelligent.
What emerges from Empis’s responses is a clear hierarchy of value. The strongest short-term returns are not in flashy consumer-facing applications, but in document-heavy internal workflows such as due diligence, lease management, and information handling. That fits with a broader pattern: the most immediate and measurable use cases in real estate tend to be those that reduce friction in information-intensive processes, particularly where speed, consistency and traceability matter. Empis’s view is that these efficiencies ultimately benefit clients not through spectacle, but through better execution: faster delivery, greater accuracy and more dependable service.
Her answers also underscore another point often underappreciated in public AI conversations: the quality of the outcome depends heavily on the quality of the underlying data. In the Portuguese residential context, valuable assets include transaction volumes, sales prices, contracted rents, new supply indicators and wider economic and demographic data. By contrast, the hardest material to operationalise is fragmented, unstructured or inconsistently formatted information: legal documents, building plans, municipal licensing data and sensitive client records. This is a reminder that the real bottleneck in enterprise AI is rarely the model alone. More often, it is the condition, accessibility and governance of the data ecosystem around it.
That matters even more as the sector moves closer to regulated, auditable AI use. The discussion highlighted disclosure, auditability, and the European AI Act timeline as central issues for property marketing and content production, particularly regarding AI-edited visuals and AI-generated copy. Empis’s response is measured and direct: transparency has to be fundamental. In her framing, any content materially generated or altered by AI should be clearly identified, with human review and auditable internal processes forming part of a wider change-management framework. This is not merely a legalistic concern. It goes directly to credibility.
In residential real estate, especially, where transactions are emotional as well as financial, client trust remains the most valuable asset of all.
The same logic extends to visibility inside generative search and answer engines. This shift also highlights the growing importance of GEO, or Generative Engine Optimisation, as conversational interfaces begin to shape how users discover properties and brands. Empis’s advice here is strikingly disciplined. Rather than chasing gimmicks, she points to the need for high-quality, authoritative content; structured data that machines can easily parse; and a stronger emphasis on factual accuracy, attribution, and lifestyle relevance. In practice, that means content that does more than list features. It should explain how residential developments support lived experience, including wellness, green space and personalisation.
For brands hoping to appear inside AI-generated answers without compromising compliance, this is a sensible roadmap: become more useful, more structured and more trustworthy.
Just as importantly, Empis does not present generative AI as a universal substitute for human judgment. On the contrary, some of the most valuable functions in real estate are likely to remain least affected over the next three years. She identifies high-stakes negotiation, strategic client advisory and physical site or property management as areas where human presence, contextual judgement, emotional intelligence and relationship-building remain central. This is one of the most credible parts of her argument. AI may compress administrative work, accelerate analysis and assist service layers, but the final act of trust in real estate, whether in negotiation, investment advice or physical oversight, is still deeply human.
That creates an interesting tension at the heart of the sector. The parts of the business most ready for AI are often those furthest from the client’s emotional field of vision: internal documents, process management, data synthesis, search support and content preparation. Yet the functions least susceptible to automation are often the ones clients value most intensely at moments of consequence: negotiation, advisory and physical decision-making. The firms that navigate this transition well are therefore unlikely to be those that simply automate the most. They will be the ones who know where automation strengthens service, where human judgment must remain in control, and how to combine both without confusion or loss of accountability.
Photo: Maria Empis by JLL Portugal
Interview: Maria Empis, Co-Head of Residential, JLL Portugal
Where do you see the strongest client ROI today: document-heavy workflows (leasing/due diligence), customer support, or marketing/distribution?
The analysis indicates that the real estate industry is currently focused on optimising internal efficiency as a foundational step.
The JLL “Global Real Estate Outlook 2026” highlights that companies are piloting an average of five AI use cases simultaneously, including data workflows, portfolio optimisation, and market analysis, which aligns with document management and due diligence. The same analysis reveals that only 5% of companies have achieved most of their goals in these projects, suggesting that the full ROI has yet to be realised.
From our perspective, the greatest short-term ROI potential lies in optimising these internal, document-heavy workflows. By automating and streamlining processes like due diligence and lease management, we create efficiencies that indirectly benefit the client through faster, more accurate service. Combined with our strategic marketing and dedicated customer support, this operational excellence is the foundation for scaling our business and delivering faster, more accurate service and consistently superior value to our clients.
Which Portuguese data assets are proving most valuable for internal LLM (“GPT”) use cases, and which are hardest to operationalise?
The most valuable assets would be transaction volumes, sales prices, contracted rents, and new supply figures, as detailed in our “JLL Market 360 REPORT 2025-2026”. Economic and demographic data from sources like INE and the Bank of Portugal are also critical.
The most difficult to operate is invariably unstructured or fragmented data. This includes complex legal documents, building plans in diverse formats, and municipal licensing data, which often lack a standardised format. Privacy is another significant barrier, especially when handling sensitive client information in the residential sector.
How are you thinking about disclosure and auditability for AI-edited images and AI-generated copy, given the AI Act timeline and “responsible use” standards?
The JLL “Global Real Estate Outlook 2026” mentions the EU’s AI Act in the context of data sovereignty, demonstrating that regulatory compliance is on our strategic radar.
Our approach must be guided by a balance of innovation and trust. In our view, transparency will be fundamental. Any content generated or significantly altered by AI, whether it’s a project image or a property description, should be clearly identified as such. Internally, we will need to develop a change management framework for AI, as suggested in the “Global Real Estate Outlook 2026,” which would include clear guidelines, human review processes, and auditable records. Client trust is our most valuable asset, and responsible use of technology is the only way to preserve it.
On GEO: what would you advise a residential brand to do in the next 6–12 months to improve visibility inside generative answers without harming compliance and trust?
We would be focusing on creating high-quality, authoritative content that can serve as source material for these AI models, such as Create “Experience-Focused” Content: The “JLL Consumer Experience Survey” shows that consumers value wellness, green spaces, and personalisation. The brand should create content that highlights how its residential projects incorporate these elements, moving beyond property features to focus on lifestyle, Develop Structured Data: Ensure all property listings are marked up with structured data (schema) so that AI models can easily parse details like price, location, size, and amenities and Prioritize Factual Accuracy and Attribution: All content must be factual, compliant, and clearly attributed. This builds authority and trust, making it more likely that AI models will cite our content as a reliable source.
Which real estate functions do you believe will remain least affected by generative AI over the next three years, and why?
While generative AI will transform many aspects of our industry, we believe the functions that rely most heavily on complex human interaction, strategic negotiation, and physical presence will be the least affected.
The reports consistently emphasise that “experience” is a key value driver, which includes human interaction and the atmosphere of space. Therefore, the following functions will remain fundamentally human-centric:
High-Stakes Negotiation and Deal Closing: The final stages of closing complex, high-value residential or commercial transactions require nuance, emotional intelligence, and relationship-building skills that AI cannot replicate.
Strategic Client Advisory: Providing bespoke advice to a high-net-worth individual or an institutional investor on their portfolio strategy involves understanding subtle goals and risk appetites, which is a deeply consultative and trust-based process.
Physical Site and Property Management: While AI can optimise operations, the physical inspection of properties, oversight of construction, and management of on-site staff require a human presence and hands-on expertise.
These functions are about judgment and relationships, not just data processing. AI will become a powerful tool to support these roles, but it will not replace the core human element that defines them.
Where Value Will Be Created Next
The real estate sector is entering a period in which operational intelligence may matter just as much as market intelligence. Supply constraints, regulation, affordability pressures, and construction costs will remain central to the industry’s future, but the systems through which firms process information, communicate with clients, and build trust are also changing. The gap identified at the Summit was therefore not simply the absence of AI from one panel discussion. It was the wider lag between how quickly technology is advancing and how slowly many property conversations still frame strategic responses.
Maria Empis’s answers offer a useful corrective to that lag. They suggest that the next meaningful gains will come less from spectacle than from discipline: better-structured data, better internal workflows, clearer disclosure, stronger governance, and content that earns visibility by being genuinely authoritative. At the same time, they remind us that not every aspect of real estate value can or should be automated.
The industry is built not only on information, but on confidence, judgement and trust.
That is why generative AI in property should not be understood as a replacement story. It is a redistribution story. Some tasks will become faster, cheaper and more scalable. Some channels will become more conversational. Some internal systems will become markedly more intelligent. But the decisive competitive edge will still belong to organisations that can combine technological capability with human credibility.
In that sense, the future of residential real estate may not belong to those who use the most AI. It may belong to those who use it with the clearest sense of where efficiency ends, and responsibility begins.