Ariel Shtarkman, managing partner, Undivided Ventures; Rui Hua Chang, managing director, business management and investment, ESR; Charles Whiteley, vice president, global digital leader, digital strategy and AI execution lead, AECOM; Raymond Kwok, senior director and head of AI and data intelligence - properties, Nan Fung Development Limited; and Nils Pihl, CEO and founder, Auki Labs, speaking at the 2025 ULI Asia Pacific Summit in Hong Kong.
(ULI)
Artificial intelligence—AI—is quickly reshaping the real estate landscape, whether by redefining the very concept of value creation or by transforming global infrastructure demands. At the 2025 ULI Asia Pacific Summit, a distinguished panel of industry leaders convened to dissect what AI means for the sector and how organizations can harness its potential. Their debate spanned the frenetic growth of data centers, the journey of AI adoption, and seismic shifts afoot for the built environment and the workforce.
The panel comprised moderator Ariel Shtarkman, managing partner of Undivided Ventures; Rui Hua Chang, managing director, business management and investment at ESR; Raymond Kwok, senior director and head of AI and data intelligence—properties at Nan Fung Group; Charles Whiteley, vice president, global digital leader, digital strategy and AI execution lead at AECOM; and Nils Pihl, CEO and founder of Auki Labs.
Chang set the tone by highlighting an explosive spike in data center demand, fueled by the AI “hype curve.” As infrastructure providers, ESR and its peers are racing to accommodate increasingly sophisticated technical requirements from tenants—requirements that have changed dramatically in the just a single year. Yet Chang posed a crucial question for the industry: “Are we heading for an oversupply?” The historical precedent is sobering; commercial real estate has undergone similar cycles in offices and shopping malls, where an initial frenzy gave way to a glut and subsequent obsolescence.
With AI’s pace of innovation, though, the risks are amplified. A data center built to the highest AI standards in 2025 might be rendered obsolete by 2026. The challenge for real estate developers is not only to anticipate the crest of the demand wave but also to future-proof physical assets so they remain relevant while technologies—and tenant expectations—leap ahead.
From digital assistants to human-AI teams
As the discussion moved beyond infrastructure, Kwok mapped out the journey of AI adoption within organizations. What began as a novelty—retrieving information from vast data sets—has evolved into far more complex territory. Today, firms are experimenting with human-agent teams, on which AI becomes a “digital colleague.” The latest predictions, notably from Microsoft’s Work Trend Index Annual Report 2025, suggest that in the next two to five years, every employee could become a boss of autonomous digital agents, fundamentally altering how businesses operate.
This transformation, Kwok argued, is not about bolting on another tool but instead about re-engineering workflows and business processes themselves. “AI is an opportunity to rethink how we fundamentally work,” he said. As companies move through this journey, human oversight and adaptability remain indispensable: the models evolve constantly, and expectations must be managed accordingly.
Whiteley described AECOM’s pragmatic three-phase blueprint for AI integration at scale:
Internal efficiency: Automating back-office functions—human resources, procurement, and especially proposal writing, where AECOM’s in-house “Oscar” tool cuts request for proposal response times by as much as 80 percent.
Reimagining delivery: Applying generative AI in design and engineering to let smaller teams handle greater workloads and boost deliverable quality. Whiteley cited a striking example, in which AECOM partnered with Norway’s Consigli to use AI-driven platforms that deliver schematic designs within a 3–5 percent margin of error, saving as much as 65 percent of worker-hours in early-stage projects.
New revenue streams: Although details remain closely guarded, AECOM is already exploring how AI can underpin entirely new lines of business.
Across the panel, concrete examples demonstrated the productivity and precision gains already emerging from AI adoption:
Property valuation: Nan Fung Group used a new ChatGPT model for property valuation, achieving results within 5–10 percent of their traditional cash-flow models. The company also slashed research and analysis time from weeks to hours in underwriting.
Design automation: AECOM’s pilots show that AI can now generate and iterate schematic designs with high accuracy, provided that human experts remain in the loop to verify compliance and context.
Retail and building management: Pihl presented a leap into physical space. Using AI-powered glasses and computer vision, Auki Lab’s platform detects empty retail shelves in real time and autonomously triggers replenishment. In Hong Kong—a city with more than 4,000 high-rises—drones equipped with AI now perform building inspections in hours rather than weeks, dramatically reducing costs and risk.
The consensus was nonetheless clear. AI cannot yet replace human accountability, especially when physical safety and sign-off are at stake. Human involvement is essential as the final check and ethical guardrail.
Despite such impressive progress, panelists were candid about the difficulties of scaling AI in real estate. Off-the-shelf solutions rarely fit; every function—whether development, asset management, or valuation—requires tailored models and business context. For such behemoths as AECOM, deploying AI globally means balancing internal development with external partnerships, prioritizing high-impact use cases, investing in “citizen AI” training, and establishing robust governance to control costs and change management.
AI’s blind spot
A provocative note was sounded by Pihl, who pointed out the central irony of today’s AI revolution. The technology excels at cognitive, white-collar tasks—work many people enjoy—yet struggles to automate the manual, physical labor that society is more eager to offload.
“We are failing miserably to use [AI] for any kind of manual labor, which is ironic, because manual labor is the work we don’t want to do,” Pihl observed. The next great leap, he argued, lies in making the physical world accessible to AI by using spatial computing, drones, and sensors to automate tedious tasks, from shelf-stocking to building inspections. If successful, this step could upend industries even faster than AI disrupts office work.
Looking to the 2030s, the panel predicted a radical downsizing of organizations themselves. Pihl said that “the age of big companies is largely over,” thanks to AI’s ability to supercharge the productivity of smaller teams. The implication for real estate developers is stark: “There is no point in creating office space for 500 coworkers anymore . . . . By 2030, the average company size will be radically smaller, and you’re not going to find tenants for those office spaces.” The message for the industry: Stop building for yesterday’s tenants and start planning for a leaner, smarter future.
