An ENR article reported that the agentic AI provider Trunk Tools lost its API access to Procore in September. According to ENR, Agave (a data integration platform) informed Trunk Tools and other startups that Procore’s new terms required changes to how they used its data connector. Procore says the change is due to its updating the API terms and conditions and introducing a new Developer Policy, aiming at protecting the integrity and security of all its customers’ data.
This incident illustrates the tricky relationship between AI startups and incumbents. Both benefit from each other, but the collaboration is not without caveats. Sometimes it can be an existential question for the startup.
The Trunk Tool episode also poses an essential question in construction technology: Who truly controls the data powering AI innovation?
The question of data
When generative AI became popular, several startups emerged that used AI on construction data. Their challenge was obtaining that raw material. Some startups collected and used data they acquired directly from customers. However, a practical solution for others was connecting to large design and project management platforms where customers were already present, using APIs to access their data.
Riding on top of the existing platform’s data also offers the advantage of accessing a well-established customer base. Customer acquisition, the most challenging task for a startup, becomes significantly easier this way. The large company gains from this mutual relationship, as it retains its customers who enjoy additional benefits from startups’ services.
As we’ve seen, relying on other platforms carries risks because they can change their terms of service at any time. In the worst case, a startup may lose its ability to serve its customers as it currently does.
Incumbents disrupting startups
The common belief is that startups disrupt established companies. They tend to be more innovative, move quickly, and generate enthusiasm that large corporations rarely produce. But when it comes to AI, incumbents can disrupt the startups.
AI-powered startups face a big challenge when they base their business on somebody else’s platform. Incumbents can add new features that mimic what the startup offers. When a startup creates something customers like and use, it shows that a real need exists. A large company can naturally buy the smaller one if it makes sense. But if the startup’s innovation isn’t protected or is superficial, the bigger company can copy their success.
Arvind Veluvali discussed this threat in his article titled Hard Truths from Procore Groundbreak 2025. He shared his thoughts on Procore’s Groundbreak event: “Procore’s launch of Helix, AI Agents, and their Developer Portal presents an existential threat to a generation of incredibly well funded construction tech startups. […] One might argue that Trunk Tools, Datagrid, and others have a better product. That might be the case, but distribution usually beats innovation.”
What can startups do?
So, what can startups do to avoid becoming obsolete or run over? They can patent their technology or try to create a technological moat to protect it. But, as Jeevan Kalanithi told me on the AEC Business podcast, “any moat can be defeated.”
Creating excellent user experiences has been effective for me. When users can get started quickly and have an intuitive experience instead of relying on a lengthy user’s manual, even large clients are more willing to switch. Additionally, providing good and responsive customer service is a winning approach.
Integration is essential, but don’t build your house on rented land. Owning the data-collection pipeline (e.g., sensors, IoT, job-site capture) makes you independent of a platform. Being platform-agnostic is another strategy that can make you less vulnerable.
Read the license
Many SaaS or platform agreements state that the “customer owns the customer data” or “the data you upload/input remains your property.” At the same time, these agreements frequently grant the provider the right to host, process, analyze, aggregate, and, if applicable, anonymize the data for the purpose of providing the service or improving it.
The contract may also allow the provider to set access, redistribution, sublicensing rules, or even to restrict or change access. For instance, in SaaS agreements, the provider may retain rights to “output data” or derived data, or to limit what the customer can do with the data after termination.
So, it is advisable to read the terms and conditions carefully, whether you’re a customer or a third-party service provider using customers’ data on a SaaS platform. I’m sure that, given the rapidly changing AI and security environment, SaaS companies will review and update their terms more frequently than before.
These contractual nuances determine not only present rights but also how the next generation of AI systems will interconnect.
The data-driven future
I don’t know whether investors assessing construction-AI startups have included “platform dependency risk” in due diligence. If not, I’m sure they will in the future.
I’ve previously discussed the agentic AI future where agents work together and access data sources using open standards. Technically, this is possible, but the main question is how user terms for platforms, apps, and data sources will enable it.
Anyway, AI has created an entirely new landscape where startups, incumbents, and most importantly, users of the systems must learn to navigate.
View the original article and our Inspiration here
Leave a Reply