Proptech money is moving to AI that cuts construction and property costs

Proptech funding is holding steady. The checks are going somewhere new. Crunchbase News reported in its Sept. 1 sector snapshot that real estate tech investors have become more selective and are directing more of their money to companies that use AI and other technology to make construction, property operations and real estate transactions faster and less expensive.

That is a sentence about venture capital. Read it from the desk of a general contractor's preconstruction director and it says something else: the tools most likely to get funded, supported and expanded over the next few years are the ones that promise to take hours out of estimating, submittal review and drawing coordination. One New York company already shows what that looks like in practice.

Mary Ann Azevedo profiled founder and CEO Sarah Buchner on Aug. 13 as part of a Crunchbase News series on non-tech founders raising venture money, describing her as a blue-collar worker by background who started as a teenage carpenter. A week later, Marlize van Romburgh's Aug. 21 column on interesting startup deals was, in her words, all AI, with the headline pointing to applications ranging from recycling to winning construction bids, according to Crunchbase News.

Bid leveling is the first job the AI takes over

Trunk Tools launched TrunkBid on March 26, 2026, according to Construction Business Owner, which carried the company's announcement. The agent handles subcontractor bid analysis and leveling, and it went out in beta to a select group of customers rather than as a general release.

The mechanics are straightforward. An estimator uploads the subcontractor bids and any scope-of-work documents; the agent pulls the relevant data from each, maps line items to the requirements, flags what each sub covers, excludes, prices as an alternate or leaves out entirely, and exports a structured head start to Excel. Trunk Tools says builders can save hundreds of human hours in preconstruction by skipping the manual setup and comparison.

That last figure is the company's own estimate, not a measured result from a customer. It still matters as a framing number. The announcement describes the problem it targets in concrete terms: subs write bids under severe time pressure, templates are not standardized, language is ambiguous and pricing is inconsistent, so a GC can miss a scope gap and default to the lowest number.

For a contractor whose estimators still level bids by hand in spreadsheets, that is the condition under which this tool is relevant. For one that already runs bids through a structured procurement system with consistent scope templates, the gain would be smaller, and the vendor's hours-saved claim would need testing against the firm's own baseline.

Notice where the output lands. Forbes contributor Louisa Xu observed in a 2019 piece that Excel remained the most commonly used data management tool in real estate three decades after its introduction, from small owners to large investment firms. Seven years later, the newest AI agent in construction procurement still finishes its work by writing to a spreadsheet.

Drawings are the harder problem

Trunk Tools followed TrunkBid with Cortex on June 17, 2026, reported by Bryan Gottlieb at Engineering News-Record. Cortex is an AI layer trained on construction drawings, markups and the visual conventions of the trade, built to connect drawings with specifications, RFIs, submittals, schedules and change orders. The company says it powers seven workflow agents and took four years of development alongside some of the largest general contractors in the country.

ENR framed why drawings are the sticking point. Software has to read symbols, revisions and relationships across hundreds or thousands of sheets in a single project set and tie those elements back to project records. Trunk Tools and many large contractors continue to describe that interpretation as a significant obstacle even as AI tools spread across the industry, according to ENR.

Buchner told ENR the company's models can now detect hundreds of different object types on drawings and compare drawing sets against each other. The capability ENR singled out is identifying changes that fall outside revision clouds, which it described as a common source of missed scope changes and downstream rework. Buchner cited one recent electrical issue outside a revision cloud that she estimated carried roughly $60,000 in project impact.

One example is not a benchmark. It is, however, the kind of number a project executive recognizes: a single unflagged revision on a single trade, and the change-order conversation starts at five figures. For contractors whose disputes tend to begin with a missed revision rather than a missed bid line, the drawing-comparison function is the one worth testing first.

ENR also noted the structural reason this matters. Despite years of attention on 3D models, most construction worldwide still treats documents as the single source of project truth, for reasons of practicality and legal risk. An AI that reads 2D sheets is aimed at how projects are actually run, not how they might be run.

Who has deployed it, and who else is in the field

Gilbane has rolled out earlier Trunk Tools agents across its jobsites, according to ENR. Cleveland Construction, based in Mentor, Ohio, piloted the platform before expanding its use.

Elliot Christiansen, Cleveland Construction's senior vice president of operations, told ENR that submittal review often lands on less-experienced project engineers even though it is among the more technical responsibilities in project management. That observation points to where the value shows up on a real org chart. Cleveland Construction, based in Mentor, Ohio, began piloting Trunk Tools before expanding its use, according to ENR.

Trunk Tools is not the only vendor moving this way. ENR reported that Procore has been integrating DataGrid, a recent acquisition, and its AI agents into its platform. That gives a buyer a real choice to evaluate: agents that live inside the project management platform already in use, or a standalone AI layer that connects to existing documents.

The Forbes piece from 2019 is worth one more line here. Xu pointed out that the companies founded in the first proptech wave, naming Autodesk, CoStar and Argus, still held leading market share at the time she wrote. Whether AI-native tools end up standing alone or get absorbed into those platforms is the open question a CIO signing a multi-year contract has to weigh, and the Procore move suggests incumbents intend to compete for the same workflows.

The other bet: cheaper transactions

Crunchbase's thesis has a second half beyond construction. The Wall Street Journal's Yuliya Chernova reported on Jan. 13, 2026 that venture investors were targeting startups aimed at easing the difficulty of buying a home, with models such as rent-to-own and no-commission sales. Dan Wenhold, a partner at Fifth Wall Ventures, told the Journal that affordability-focused tech was top of mind for investors, partly because of shifts in the broader policy climate.

Put the January and September reports side by side and the pattern is consistent. Investors are backing technology that lowers the cost of a building at both ends of its life: the price of putting it up and the friction of changing hands. Crunchbase News has also reported this year that proptech is one of the sectors where seed rounds of $5 million to $10 million are clustering.

For a property operations leader, the practical read is that the vendor field for cost-reducing tools is being funded at the early stage right now, which typically means more pilots on offer and more pressure to show measured results. For a construction procurement lead, it means the estimating and submittal workflows are where the first products have actually shipped.

What would confirm the read

Crunchbase's snapshot describes investors as more selective without publishing the metrics behind that judgment, so selectivity and a general pullback cannot be separated from the text alone. The stronger evidence is on the product side: two agent launches in three months from one vendor, a platform incumbent integrating an acquired AI team, and two named contractors past the pilot stage.

TrunkBid was limited to a select beta group when it was announced in March. The next signal is whether Gilbane, Cleveland Construction or the beta customers report measured hours saved once the agent opens more broadly, turning the company's 'hundreds of hours' estimate into a number a preconstruction director can put in a budget.