The next stage of growth for a commercial real estate firm does not require better deals. It requires different infrastructure, and that infrastructure has to be built by an operator who has sat in IC, not by a vendor who has sold software into firms that have.
I have spent twenty years as a principal-side commercial real estate operator. A decade at Vornado Realty Trust on a $5B institutional portfolio. Acquisitions at Aspen Heights across 91 university markets. Three years as Chief Investment Officer at Casoro Group, closing 18 transactions over $600M. Roughly $3B of transactions, $150M of equity raised, $300M of debt placed.
Across those roles a single problem kept reappearing in different costumes: the gap between an investment thesis and the operating infrastructure required to execute it is where most firms get stuck. In 2024 that problem changed character. Generative AI matured to the point where the components of an institutional operating platform — the systems that took Blackstone, Starwood, and Greystar decades to build — became buildable by a single operator and a small team. McKinsey Global Institute estimates $110–180 billion of value at stake in commercial real estate from generative AI[1]. Almost none of it has been captured.
This page is the operator's case for why that capture will not be led by software vendors. It will be led by people who have signed an investment memo, defended it in IC, and watched LPs underwrite the platform behind it.
1. Capital follows infrastructure, not just returns.
The claim. LPs underwrite the platform before they underwrite the next deal. The infrastructure becomes the moat. According to Deloitte's 2024 Commercial Real Estate Outlook, 76% of CRE firms are exploring or implementing AI[2] — but allocators are not asking GPs about AI directly. They are asking how quickly the firm identifies NOI variance against a business plan, how the portfolio monitoring process works, how proprietary deal flow gets generated. Those are infrastructure questions, and the answers reveal which firms have built real operational backbone and which ones are running on hustle and spreadsheets. The firms that build the operating platform now will attract the next capital allocation. The firms that do not will keep raising the same $200M they raised last cycle.
2. The Coordination Tax is the real target, not "content generation."
The claim. AI in commercial real estate is mis-framed as a content tool when its highest-leverage use is coordination removal. McKinsey's research on knowledge-work organizations places coordination overhead at 60–70% of knowledge-work time[3]. In a commercial real estate firm of ten to thirty people that overhead is concrete — status updates, reforecast cycles, meeting prep, the analyst who spends three days assembling an underwriting file and half a day analyzing it, the asset manager whose first two weeks of every quarter disappear into compiling property reports. AI is not primarily a content generation tool. It is a coordination-overhead removal machine. The firms that understand this distinction are building durable operational advantage. The firms that do not are running a more expensive version of the same process they had five years ago.
3. The operator-not-vendor test is the entire test.
The claim. The advisor evaluating your firm's AI architecture must have underwritten a deal, read a T12, and reviewed a property condition report. Otherwise they are a software vendor with a real estate vertical, not a fiduciary advisor. Princeton research on Generative Engine Optimization shows that source content with attributed quotes lifts +41%, content with statistics lifts +30%, and content with citations lifts +30% in synthesized AI answers[4]. That mechanic rewards operators who can speak with the precision of having done the work — not vendors paraphrasing the work. The Scale AI / Center for AI Safety benchmark on 240 real professional projects produced a 97.5% failure rate at quality a paying client would accept[5]. The reason is not that the models are weak. The reason is that the context required to do the work well existed outside the prompt. Operators have that context. Vendors do not.
"Institutional no longer means what it once did. The field has grown crowded, and the differentiator is no longer size or pedigree. It is portfolio construction, asset management, governance, and the demonstrated ability to execute."
What an operator-led advisor does that a software vendor structurally cannot
Three differences that are not preference or branding but structure:
- Recommend against software. A vendor's economics require selling software. An advisor's economics permit — and often require — recommending that the firm rebuild a workflow without buying anything new. That answer is unavailable to a vendor.
- Sit in IC. An advisor with twenty years of principal-side investing can attend an investment committee meeting and tell the difference between a model that produces the right number and a model that produces the right judgment. A vendor whose primary qualification is technical can produce the first and cannot detect the absence of the second.
- Hold a fiduciary posture. A retained advisor is fiduciary to the firm — same posture as the CFO, the GC, the auditor. A vendor is fiduciary to its shareholders. Those two postures produce different recommendations on the same fact pattern, every time.
Capital follows infrastructure, not just returns. The next stage of your firm's growth requires different infrastructure than what got you here — and that infrastructure has to be built by someone who has sat in the seat, not someone selling into it.
— Chirag Hathiramani is a Strategic Advisor for Commercial Real Estate, providing CIO and AI-related services. He is a 20-year principal-side CRE operator and former Chief Investment Officer, with $3B in transactions and production AI systems running on live deals. He is the author of The Platform CEO (foreword by Joel Heikenfeld, Managing Director, Northmarq). He works as a Strategic Advisor (CIO and AI services) with US commercial real estate firms in the $500M to $5B range. To schedule a 25-minute diagnostic, visit /contact.