The commercial real estate industry's multimillion-dollar bet on generative AI has produced a sobering outcome: near-zero measurable business impact. According to Brian Connolly, founder and CEO of Feasibly, an AI platform for market studies and financial feasibility analysis, the fundamental problem isn't the technology itself but rather a widespread misapplication of generic tools to highly specialized workflows. Connolly, who spent years conducting feasibility work at Victus Advisors before launching Feasibly, argues that most CRE organizations started with the wrong instrument—enterprise chatbots and copilots that lack the structure and domain expertise required for nuanced deal evaluation.
The evidence supporting Connolly's thesis is striking. A 2025 MIT study found that 95% of organizations reported no measurable profit-and-loss impact from their generative AI investments, despite enterprise spending reaching $30 billion to $40 billion on the technology. These chat-interface systems, Connolly explained in a recent interview, regularly provide different answers to different employees asking the same question, introducing uncertainty rather than driving efficiency. The technology works, he contends, but commercial real estate companies have applied it incorrectly to workflows that demand consistency and precision.
The disconnect between broad AI models and specific business problems has created a valley of disillusionment across the industry. JLL's survey of over 500 senior decision-makers last year revealed that while 88% of investors, owners and landlords are running AI pilots, 60% report being unprepared to effectively implement AI into business workflows. Without structure, domain expertise and clearly defined processes, Connolly maintains, AI exists only as a novelty. Many pilot programs stall in the experimentation phase, never graduating to production deployment that could justify the initial capital outlay.
Connolly advocates for a fundamental shift from chatbots to what he terms agentic AI—systems designed to execute specific jobs rather than simply answer questions. These systems are structured, goal-oriented and capable of completing multi-step workflows with consistency and minimal human oversight. Unlike chatbots that aimlessly search through sprawling corporate datasets, agentic AI systems function as precision instruments, directed to find key information in precise locations and instructed on how to apply it for specific business use cases. The distinction, Connolly offered, is between an intern searching through files and a trained specialist who knows where to look, what matters and how to act on findings.
That distinction carries particular weight in commercial real estate, where consistency and accuracy underpin capital allocation decisions worth millions of dollars. When evaluating deals, underwriting risk or making investment commitments, navigating variability represents a critical step toward project success. Chatbots create additional uncertainty by generating responses dynamically, whereas agentic systems remove variability by following structured processes. This architectural difference makes purpose-built systems far more reliable for real-world applications where precision determines whether a development pencils out or a renovation strategy proves viable.
Market studies and financial feasibility analysis represent the most immediate proving ground for this approach. Traditional feasibility studies for commercial real estate projects can cost $50,000 or more and require months of human analyst time, creating an insurmountable roadblock for many executive decision makers. Developers, investors and lenders rely on market feasibility analysis to assess whether projects make financial sense, but the slow and expensive nature of conventional processes limits how many opportunities can be evaluated. Agentic AI has fundamentally transformed this workflow, reducing both time and cost by up to 90% according to Connolly.
The efficiency gains stem from orchestrating multiple specialized AI agents, each responsible for a discrete part of the workflow. In Feasibly's system, tasks such as data retrieval, market validation, benchmarking, forecasting and narrative synthesis each have a dedicated, customized AI agent. By decomposing what was previously a highly manual process into defined steps and assigning each step to a purpose-built agent, the workflow transitions from manual to automated. Connolly noted that the process isn't entirely hands-off, acknowledging that human oversight remains a component of quality control.
The broader implication extends beyond cost reduction to deal velocity and risk mitigation. By providing reliable insights at critical early-stage milestones, agentic systems accelerate the path from initial concept to viable project. Developers can evaluate more opportunities in the same timeframe, investors can conduct preliminary screening at lower cost, and lenders can assess risk profiles before committing substantial underwriting resources. The technology enables a different cadence of decision-making, where feasibility questions that once took months can be addressed in days or weeks, fundamentally altering the economics of deal evaluation in capital-constrained markets.