Originally published by Thesis Driven. Republished here with permission. Read the original article here.

For a long time, services were a boring backwater: thrashy, low-margin work that perhaps needed to get done in service of something worthwhile (such as selling a SaaS subscription) but best avoided for any company that cared about growing quickly and at high margins.

Over the past two years, the tables have turned. Companies that would've sold per-seat software subscriptions five years ago are now rolling out service packages, powered by their own in-house software.
But not every attempt to vertically integrate looks the same. While some companies organically expand into services, others embrace a more aggressive M&A strategy. And venture investors are following the shift, increasingly seeking out companies with a services strategy, a situation that would’ve been incomprehensible just five years ago. Today's letter will explore the vertical integration trend in real estate tech: why it's happening now, how differently companies are approaching it, and what it means for the industry's investors and entrepreneurs.

The End of the SaaS Premium

SaaS was long the gold standard. Services, in contrast, were seen as lumpy and unpredictable. That distinction showed up directly in the multiples each business model commanded in the public markets, with investors paying a premium for steady SaaS income.
But the SaaS premium is not what it once was. SaaS EBITDA multiples (9.7–12.2x) now overlap neatly with IT services EBITDA multiples (9.8–10.2x public, ~12x on private deals). While it’s not possible to run a perfect comparison between very different companies with very different models, the SaaS premium has clearly deflated over the past twelve months.

Services businesses, on the other hand, are now en vogue.

Take Pulley: it’s a software platform using AI to help commercial real estate firms submit permitting documents. But rather than sell it directly to architecture firms as a software tool, Pulley offers it to developers directly as a service.

"[Pulley] was always this fully managed service," explains Charlie Jacobsen, co-founder of Pulley. "That's what it takes to get great permitting outcomes for complex commercial projects. And it's also what the market demands."

The service model, of course, demands a different staffing structure; Pulley has more than 50 licensed architects on its team. "If we were a standalone architecture firm, we'd be in the top several hundred architecture firms in the country," said Jacobsen.

Pulley isn't the only software company hiring architects. We wrote about the schism among automated feasibility assessment companies last year: while several like Algoma and TestFit are taking a more conventional SaaS approach, others, such as Cove and Cedar, are choosing to build vertically integrated architecture firms. While proprietary software is still key to their value proposition, they are charging services fees comparable to architecture firms rather than per-seat software fees.

"The problem isn't just the software, it's how people are using the software," explained Sandeep Ahuja, CEO of Cove, to me last year. "Tools can be a band-aid, but they're not going to solve the problem."
In part, this shift is driven by tools getting far easier and cheaper to build. Ten years ago, creating software was far more difficult than it is today, requiring a company's entire focus and attention to do it well. Today, for many software development tasks, the actual engineering is trivial, making it easier for a company to DIY the "AI-enabled thing" that their services business runs on.

The rise of AI coding tools has also shifted the hard part from engineering to UX and product. And getting those right is often easier when the people using the software sit in-house rather than within clients' organizations. The "design partner" model, inviting a few early clients to build a product alongside you, was best practice in software product development for many years, but it’s increasingly irrelevant in a services-first world.

Now startups are their own design partners.

But the vertical integration trend doesn't stop at companies building their own services businesses. Some have taken it a step further by buying services companies: a tech-enabled rollup model.
KP Reddy is one entrepreneur who has embraced the rollup approach. His startup, Zero RFI, provides owners' rep services to developers tackling construction projects. Since starting the company in March 2026, Zero RFI has acquired several construction management firms, including Brookwood Group and Buildingworks.

"The plan is to roll up as many owner's rep firms as possible," said Reddy. From there, "we transform the companies we acquire and spin off net-new, AI-native services from them," turning them from "10-point margin businesses to 60-point margin businesses."

As we'll see, the rollup approach is a spectrum of build-vs-buy that depends on an operator's capitalization, market, and desired speed. And there are pros and cons of each path.

A Solution for Hard Problems

For some companies, moving away from a sell-the-technology approach offers promise in solving harder technical problems.

Consider LITA, a company building an orchestration layer for robots operating in hotels. Founder Batis Samadian saw a disconnect between the reality of single-purpose robots (many of which were effective and capable at basic cleaning and delivery tasks) and the lack of actual utilization in hotel settings.
One reason for that gap is a lack of effective orchestration: the software that connects the robots' brains to the building systems. For instance, delivery robots must be able to coordinate with the elevator system to call the elevator and direct it to the correct floor. So Samadian decided to build the orchestration system for robots in hospitality settings.

Five years ago, such a company would've tried to sell orchestration software directly to hotel management companies, and likely struggled mightily in that effort. Robotics, after all, faces major challenges in real-world settings that have limited its adoption. And perhaps the biggest challenge is that on-site teams (at best) don't have experience dealing with robots and (at worst) are actively hostile to their presence on site.
“You kind of face this problem of long sales cycles meets IT training tall orders,” explained Samadian. “And a serious reality is you are often facing potential sabotage by the team hired by that general manager or that hotel management company.”

If you're selling a robot to a hotel, you're not just selling technology, you're selling cultural change. And selling cultural change is really hard.

And while hotel owners were eager to take advantage of robotics' promise to cut operating expenses, operators are often far more reticent. “Owners were excited about drastic opex cuts, not slight workflow optimizations,” noted Samadian. But after the owners referred LITA to their operators, the engagements would get whittled down to minor workflow optimizations, proposals that ended up adding cost to the hotels and reducing NOI since there was no commensurate reduction in headcount.

So Samadian instead chose to offer management services himself, forming a partnership with AINA Hospitality, a hotel management company. “If we just sell the software, we're going to be stuck maximum making like $200,000 per property,” said Samadian. “Hotel management agreement actually can make one and a half to three and a half million, depending on the asset size, region, and city.”

Of course, LITA doesn't own its management partner here; rather, they're sold as a bundle. Whether LITA scales by deepening its AINA partnership, buying the company outright, or packaging its technology with multiple "preferred" management companies is the call Samadian will have to make as the business grows.
Parking operator Metropolis was an early example of using vertical integration to bring tech to a knotty real-world problem. Like LITA, Metropolis faced a market with tech-curious owners but a resistant operator layer. "The operators didn't have the technical know-how or the willingness to deploy technology at scale," explained Metropolis co-founder and CEO Alex Israel. "So we realized we had to vertically integrate."
After buying its first parking operator in 2022, the ninth largest in the US, Metropolis proved the ability of technology to add value in parking: the company's computer vision technology allows Metropolis members to enter and exit seamlessly without fumbling around with tickets, credit cards, and pay stations. The result is driver preference (users familiar with the system will seek out Metropolis garages) as well as reduced leakage, leading to an average NOI bump of 20% for real estate partners according to the company.
In 2024, they raised $1.8 billion, the majority of which was used to acquire the publicly traded, 100-plus-year-old parking operator SP+. Today, Metropolis operates the largest physical recognition network in the United States, with over 4,500 locations.

A Capital Shift

The shift toward services would not be possible if venture capital didn't get behind the idea. All these companies, after all, are venture-backed: Pulley by Founders Fund and Y Combinator, Cove by Coatue and others, and Metropolis by Slow Ventures, Dragoneer, and Zigg.

The vertical integration trend has also reached the earliest stages of investing. Incubator Stackpoint, which has built businesses across the real estate and construction industries, has embraced the AI-enabled services model. Two of their most recent incubations have a services component.

"It's really the same thing we've been doing with building vertical AI companies in these complex industries," said Adam Pase, co-founder and General Partner at Stackpoint. "[Services are] a new tool in our toolkit that can either accelerate that or unlock entirely new opportunities that we didn't see before."
For Stackpoint, services are more than an end in themselves. Pase specifically believes that a services model can boost an AI-powered business in three ways: access to a credential, data at scale, and distribution and trust.

The credential piece is important (and often overlooked by early-stage entrepreneurs) in industries with complex regulatory regimes like insurance and architecture. "Getting regulatory approvals, licenses, or rated entity status can take two to three years from a cold start for a startup to acquire," explains Pase. "And in many cases, you can't actually unlock volume until you get that credential, and you can't get the credential until you unlock the volume. So it's sort of a chicken or egg.

"Being able to do that in months instead of years in a startup is huge."

The data side is no less important, and can, in theory, build enduring moats. "The deep transaction-level data and all the edge cases, that’s what you need to train the substrate of your models," says Pase. "Those aren't available elsewhere. They're locked within these services businesses."

Finally, the distribution from services can be an accelerant for an early-stage tech company. "There's a services firm that already serves the customers your platform needs, and they've built up relationships and trust that would take a startup six to 12 months in a traditional sales cycle to earn for each of those players," noted Pase.

But Pase is careful to distinguish tech-enabled services businesses from businesses that use services to gain an edge to sell software.

"Where some people may get in trouble in this game is when they are looking at venture multiples on improved services businesses," said Pase.

"You can look at a services business that's currently operating at 20 or 25 percent margins and say, hey, I can build internal tech that helps that services business run much better and transform that into a 60 or 70-plus percent margin business. But what you've done there, in my opinion, is you've built a much better services business. That might actually be a really good thesis for a private equity firm. Not necessarily a venture outcome."

In other words, if it's a pure-play services rollup leveraging AI, it's not for Stackpoint. But if the services can make better software, it's a more compelling opportunity.

Not every investor is as skittish about pure-play AI-powered rollups. Reddy, for instance, raised a $13.8 million seed round led by General Catalyst earlier this year, which has fueled Zero RFI's purchases to date. And this kind of deal isn't an outlier for General Catalyst, which has built an entire program around AI-powered rollups.

"Their view is that they're big investors in Anthropic, so they check the box on foundational AI, and everything else is applied AI," said Reddy.

On Discipline

This is not indicative of a shift back to the happy-go-lucky VC days of 2021. Nor is it going back to the days of a quarter century ago when venture capitalists backed airlines, restaurants, agencies, and all manner of non-tech businesses.

After all, services multiples have converged with SaaS largely because SaaS multiples have declined, not because services multiples have risen. In many ways, services remains a tougher business requiring a far more disciplined approach. Companies that raised large amounts of venture capital at high valuations on a tech vision before pivoting into services are unlikely to be saved unless those services businesses get very, very large. Metropolis is the exception, not the rule.

Services also introduce problems not faced by SaaS companies. Managing non-exempt employees in a services context is a very different task from managing software engineers and product managers. Of course, this depends a lot on the specific services context; architects act more like engineers than they do maintenance techs. But while operating Common, we dealt with more than one fistfight between site employees, and hourly workers often don't give notice before they quit: they just don't show up. It's a different animal than running a tech company.

And not every business benefits from a pivot to services. Most of the companies profiled here face incumbent services providers running on legacy technology stacks, and often, management teams resistant to change. For tech companies, the service model works if it does at least one of the following extremely well: (1) loosens the sales funnel by shifting the target to one with higher agency and more tech savvy, (2) generates a materially higher take rate, or (3) provides critical data they could not get otherwise.
Ideally, it does all three.

But there is one incontrovertibly good thing about the shift to services: it aligns entrepreneurs more clearly with the outcome they promise to clients. "Real estate firms have a line item on their P&L for a third party service where they're paying for the outcome, not the software," said Pase. "One of the benefits of looking at the services lines is that they're already paying for the outcome."

—Brad Hargreaves