Adir Levitas, CEO, Faropoint

Guest: Adir Levitas, CEO, Faropoint

Small Bay Industrial at $5B Scale - Built on Proprietary AI

 

Adir Levitas of Faropoint explains how proprietary AI platforms now process 65% of small bay industrial deal flow across the US.

 

Small bay industrial AI underwriting is no longer a differentiator on the margin - at Faropoint it is the operating architecture. The firm manages over $5 billion in assets across 550 buildings and 30 million square feet, acquiring 100-120 assets annually at roughly two properties per week. That pace is only viable because Faropoint built proprietary software - including an acquisition pipeline platform, a machine learning rent-forecasting suite called REXy, and an asset management intelligence platform - that together process approximately 65% of all small bay industrial deal flow in their target markets. The competitive result is a Bloomberg-screen-like view of what is for sale in real time, before most buyers know a property is available.

 

Key Takeaways

• 65% market coverage is a structural moat.
  Faropoint's acquisition platform captures the majority of small bay industrial deal flow in its US target markets, giving the firm real-time visibility into bid-ask spreads, cap rate differentials, and portfolio versus single-asset pricing - before competitors receive a broker email.

• REXy benchmarks rent accuracy above human underwriters.
  This AI powered suite provides market rent estimates, tenant renewal probabilities, and two-year rent growth forecasts for industrial suites across the US, with accuracy that Levitas says consistently exceeds what analysts produce manually.

• Investment professionals are now building the software.
  Two senior investment team members - not engineers - built the front ends for Faropoint's acquisition and asset management platforms using AI coding tools. The tech team supplies back-end architecture and data security; the business team drives product decisions.

• Institutional LPs have shifted from skepticism to due diligence.
  Between 2018 and 2024, AI capabilities were a conversational footnote with investors. Since late 2023, the same investors are scheduling dedicated data science team meetings with Faropoint to assess whether the firm is built for the next decade.

• The acquisition intelligence platform compresses a full deal cycle into one app. From broker email ingestion through tenant credit scoring, underwriting model, due diligence document review, and IC memo generation - the entire process runs inside a single platform, enabling a small team to handle significantly higher transaction volume.

• Autonomous vehicles and robotics are identifiable but manageable threats.
  Both trends favor last-mile urban warehouses over logistics hubs in secondary locations. Proximity to residential density - the core of Faropoint's site selection thesis - is the primary hedge against both disruption vectors.
• The compounding effect is the argument, not the individual tool.
  No single AI feature produces a detectable edge. The 250 basis point outperformance case rests on the cumulative effect of better underwriting data, faster deal processing, higher-quality due diligence, and reduced operating leverage per transaction.
  

With more than 30 years and $1.5 billion in commercial real estate transactions - including large-scale industrial investment across multiple cycles - Adam Gower has watched technology claims come and go without restructuring how deals actually get done. The conversation with Adir Levitas is a rare case of an operating firm that built the systems first and proved the thesis at scale. For sponsors wondering what AI integration looks like beyond a ChatGPT subscription, this episode is a working answer.

How Faropoint Built a 65% Market Coverage Platform for Small Bay Industrial

The small bay industrial market - warehouses between 20,000 and 100,000 square feet - comprises roughly 225,000 of the approximately 300,000 warehouse properties in the United States. Despite representing the majority of buildings, it is one of the least institutionally penetrated segments in commercial real estate. That inefficiency is the opportunity; it is also the operational problem.

Acquiring at the pace Faropoint requires - $1 billion in assets annually, across 100 to 120 discrete transactions - demands a deal flow infrastructure that email and spreadsheets cannot support. Six years ago the firm built its own acquisition pipeline software and began asking brokers to upload properties directly rather than send them individually. The platform now works with close to a thousand brokers nationally. Those who do not upload directly can email the platform, and a large language model (LLM) - a type of AI that interprets natural language - converts the submission into a structured deal record automatically.

Levitas describes the result: "We have sort of a Bloomberg screen into most of that product when it's trying to be sold in real time. What's nice about it is not all of it is being sold. So you can see the bid-ask spread of stuff that was sold or wasn't sold. You can see cap rate between markets. You can see relative value."

The 65% coverage figure means Faropoint sees the majority of properties marketed for sale in its target markets before or as they hit the broader broker network. At an acquisition pace of two properties per week, the selection advantage compounds. The firm is not bidding on everything; it is choosing among a near-complete set of available inventory. The firm is not bidding on everything; it is choosing among a near-complete set of available inventory, with full visibility into the cap rate differentials that determine relative value across submarkets.

REXy and the Machine Learning Underwriting Suite

Deal flow visibility solves one problem. Underwriting accuracy at high volume solves a different one. Faropoint addressed the second through an AI powered tech suite internally called REXy - an analogy to Zillow's Zestimate, applied to industrial lease economics rather than residential sale prices.

REXy currently runs four models. REXy for rent estimates market rent for any suite in any US industrial property; REXy for renewal calculates tenant renewal probability; REXy for sale produces a value estimate with confidence weighting; and REXy for near-term market rent forecasts rental growth by submarket over a two-year horizon. The models are trained on the proprietary data generated by Faropoint's acquisition platform - years of bid-ask, leasing, and performance data that outside vendors cannot replicate.

The practical application is a benchmarking capability that the firm treats as proof of performance, not just internal tooling. As Levitas explains, leasing above a known market benchmark is the definition of alpha: "If you lease above that market, that's your alpha. Think about if you have a market of what you can lease things to... others don't even know where the market is because maybe they're lacking the AI tools to say that, or they're lacking the data to base that on."

That framing is important for institutional LP conversations. Faropoint can show quantitatively whether it is leasing above or below the modeled market in any submarket at any point in time. The alternative - human judgment based on broker feedback - cannot produce an equivalent audit trail. Faropoint uses it as an internal performance benchmark - proof that they leased above market, not just a claim. AI-powered due diligence in commercial real estate is reshaping how sponsors verify that kind of assertion.

Dissolving the Technology Silo: Investment Teams Building Their Own Software

The organizational change Faropoint made may be more significant than any individual tool it built. The traditional model - a separate technology department producing quarterly feature releases for an investment team that could not influence the product - is gone. In its place is a structure where investment professionals build production software and the technology team provides infrastructure.

Two examples illustrate the shift. Yannai Gordon, who leads Faropoint's Florida market, built the front end for the asset management intelligence platform. Mike Tang built the front end for the acquisition platform. Neither is an engineer. Both are investment professionals who used AI coding tools - specifically Claude Code - to build functional software with business logic they understood directly from their own deal work.

Levitas describes the precondition for this to work: "I took over the management of the tech team... I forced myself to understand what is the architecture data scientists are doing. What is the back end when they say back end, what does that mean? When they say DevOps, what is DevOps? I had to go into it." Leadership fluency in both domains was required before the organizational barrier could come down.

The firm also built an internal tool that is a secured, Claude Code-based environment connected to all internal data systems and external data feeds. Any employee can use this tool to build a purpose-specific application without writing raw code or accessing sensitive data outside controlled parameters. Levitas gives the example of a staff member building an investor tour app in minutes: properties within a five-mile radius, three-hour drive time, one-pager and business plan per asset. The tech team supplies infrastructure. The business team drives product. The silo is gone - and for CRE sponsors looking to replicate that model, AI integration across the deal lifecycle is the starting point.

AI Risk Factors in Small Bay Industrial: Autonomous Vehicles, Robotics, and Tenant Concentration

Levitas identifies three AI-related risks to the small bay industrial thesis and addresses each specifically.

Autonomous vehicles are the most commonly cited concern in last-mile logistics. His analysis: the risk concentrates in mid-continental distribution hubs, not urban last-mile facilities. "When somebody has to ride from Chicago to LA, they need to stop somewhere in the middle. But if your facility is a last-mile facility to deliver to a million people in a 20-minute drive, that has nothing to do with that." Urban density proximity - the core site selection criterion for small bay - is a structural hedge.

Robotics and vertical stacking present a different issue. Most small bay tenants do not stack goods because inventory turns quickly. Robotics-driven stacking is more relevant to larger distribution facilities. However, Faropoint is already working with a Fortune 500 company on a roof raise from 16 to 32 feet clear height - suggesting that even last-mile tenants will begin requesting robotics-compatible configurations in well-located assets.

The third risk is less often discussed: AI-driven consolidation among large tenants, which could reduce tenant diversity. The hedge is tenant base diversification across assembly, manufacturing, distribution, third-party logistics (3PL), and global logistics. A building suitable for all of those uses is insulated from any single industry's contraction. Levitas frames the underlying principle as Bezos's question: "What in your business is probably not going to change in the next ten years? Bet on that."

Frequently Asked Questions

What is small bay industrial real estate and why is it harder to scale than large-format industrial?

Small bay industrial refers to warehouse properties between 20,000 and 100,000 square feet, typically occupied by local and regional businesses for distribution, light manufacturing, or storage. The management burden per square foot is similar to large-format assets - an asset manager, property manager, and accountant are required regardless of building size - but the revenue per building is far smaller. Achieving institutional returns requires significant portfolio scale, which in turn requires acquiring a high volume of individual assets each year. That volume is operationally infeasible without process automation and purpose-built deal management software.

How does Faropoint's REXy platform differ from third-party rent comp data?

REXy is trained on Faropoint's proprietary transaction and leasing history across its national portfolio, supplemented by the deal flow data generated through its acquisition platform. Third-party comp data aggregates public and brokered information, which lags actual market activity and does not account for building-specific or submarket-specific nuance at the granularity small bay requires. REXy also produces renewal probability and near-term rent growth forecasts, not just point-in-time estimates - which makes it an underwriting tool rather than a reference database. The key distinction is that Faropoint uses REXy as an internal performance benchmark, not a market survey.

How are institutional LPs evaluating AI capabilities in CRE managers today?

According to Levitas, the conversation shifted materially in late 2023 and accelerated through 2024. Previously, AI integration was treated as a novelty - interesting to mention but not weighted in manager selection. Today, some institutional investors are scheduling dedicated data science team meetings with GPs to assess technology depth. The framing has moved from "do you use AI" to "are you structurally built to operate competitively over the next decade." GPs who cannot demonstrate an integrated AI capability - not just a subscription to off-the-shelf tools - are increasingly exposed in that conversation.

What does it actually take for a CRE firm to integrate AI at the operating level rather than using it for task automation?

Levitas's answer is organizational before it is technical. The preconditions are: senior leadership that understands enough of the technology to break down the barrier between investment and engineering teams; a culture that measures shared outcomes rather than departmental deliverables; and tools that let business-context professionals build directly rather than relay requirements to a technology team. He is explicit that buying an AI tool and assigning it to a tech silo will not produce competitive advantage. The firms that get there are the ones where partners and senior investment staff change their own workflows, not just their team's. 

Related Resources

Tier 1: For a broader framework on how AI is restructuring deal sourcing and underwriting across CRE asset classes, see how AI is reshaping the CRE deal cycle on GowerCrowd.

Tier 2: If you are a CRE sponsor building or evaluating an AI-integrated operating model, the GowerCrowd AI Accelerator Program is a structured curriculum for exactly that. 

Adam Gower, Ph.D. is the founder of GowerCrowd and one of the most experienced practitioners in commercial real estate capital formation. With more than 30 years and $1.5 billion in transactional experience - including serving as President of a Universal Studios development division overseeing $400 million in projects across Asia Pacific - Adam now helps CRE sponsors build AI-powered systems for investor acquisition, deal management, and capital deployment. His clients collectively manage over $45 billion in assets under management. gowercrowd.com