Who owns your construction data?

Three ways contractors lose access

None of these involve a vendor claiming to own your data. In every case the contract says the data is yours. The access goes anyway.

1. Someone else owns the account

The most common version has nothing to do with the software vendor at all. A general contractor buys a platform to hold their project history. An owner then mandates that the job runs in the owner’s account instead. The GC’s records now live somewhere they do not control, and access can be withdrawn when the project closes.

Procore describes this pattern in their own writing, which is a stronger admission than any competitor could make on their behalf. Procore

“For so many subs, when the project closes, the digital door closes with it.”

Procore

The same article says that subcontractors in this position “were simply renting access” to their own project history, because the general contractor controlled the platform, the documentation and ultimately the archive. Procore’s advice is to mirror records into a system you control, and to negotiate export rights at closeout with the same seriousness as financial retainage. Procore

Contractors describe the same thing happening to them in Procore’s own community forum, where one general contractor put it plainly: “We purchased Procore so we can house the historical data and now that’s not the case.” Procore Community

2. An export exists, but it is not usable

A vendor can answer “yes, you can export your data” truthfully while the export is close to useless for the thing you need it for, which is usually moving into another system or defending a claim years later.

Procore is a fair example because their documentation is public and specific. Procore does provide extracts. Procore’s own support documentation also states that most items come out as PDFs, that the Procore Extracts application covers only projects set to Active, and that a Data Extracts 2.0 extract is available for download for 30 days. Procore support Procore support

A PDF of an RFI log is a record you can read. It is not a record you can import, query or reconcile against a schedule. The question to ask is not whether an export exists but what shape it comes out in.

Subcontractors feel this hardest, because they rarely control the platform in the first place. One industry write-up, from January 2024 and so older than the rest of the sources here, found that the systems general contractors use often lack an easy export and that what does come out is not always usable, leaving subcontractors without their own project information at closeout. SMRTBLD

3. Your data trains a model you do not control

This one is new, and it is moving quickly. Construction platforms are now competing on whether customer project data can be used to train their AI models, which makes data-use clauses a live negotiating point rather than boilerplate. MarketScale

The commercial logic is straightforward: a platform that can aggregate schedules, cost codes, subcontractor performance and RFI logs across thousands of projects builds better models than one restricted to your data alone. The value created is real. The question is who owns it, and whether it survives your leaving. MarketScale

Buyers have started screening for this. One market signal published in August 2026 found construction firms favouring vendors that offer portable data formats and treating proprietary lock-in as a disqualifying risk during vendor selection — a shift, in its words, from passive acceptance to active negotiation over data ownership terms. Quettor

What to ask a vendor before you sign

These are the questions worth asking in a sales call. Each one has an answer that should reassure you and an answer that should not. If your current vendor answers all of these well, the honest conclusion is that you are probably fine and do not need to move.

1. Can I export every record type myself, without asking support?

   What a good answer sounds like: A self-service export in the product, covering every module you use. A good answer names the formats. A weak answer is “contact your account manager”, which means the export is a favour rather than a feature.

2. What format does the export come out in?

   What a good answer sounds like: CSV, JSON or an API for structured records. PDFs are fine for documents and drawings and near-useless for anything you need to import. Ask specifically about the records you would need to reconstruct a job: RFIs, submittals, change orders, cost codes.

3. Is there an API, and is it included in my plan?

   What a good answer sounds like: A documented read-and-write API available on the plan you are actually buying. An API that exists but sits behind an enterprise tier is not available to you.

4. What happens to my data if I stop paying?

   What a good answer sounds like: A stated window in which you keep export access after cancellation, written into the contract. “Your account is deactivated immediately” means you must export before you cancel, which is worth knowing in advance rather than discovering.

5. If a client or GC owns the account, what do I keep?

   What a good answer sounds like: A clear answer about whose records are whose, and a route for you to retain a copy of your own work. If the answer is that it is entirely up to the account owner, your project history belongs to your customer.

6. Is my data used to train your AI models?

   What a good answer sounds like: A direct yes or no, and if yes, whether you can opt out and what happens to model improvements derived from your data. Vagueness here is itself an answer.

7. Which subprocessors see my data, and where are they?

   What a good answer sounds like: A published list, naming each third party and what it does. A vendor that cannot tell you where your data goes cannot promise you what happens to it.

8. Can I get my data out at project closeout, not just at contract end?

   What a good answer sounds like: Export that works per project, at the point a job finishes. This matters most for subcontractors, whose relationship with a platform often ends when the job does rather than when a subscription lapses.

What the contract should say

Four things are worth having in writing. None of them is unusual to ask for, and a vendor who resists all four is telling you something.

• Ownership, stated explicitly. Your data is yours, said plainly, rather than inferred from the absence of a claim to the contrary.
• A defined exit path. What you can export, in what format, and how long you keep access after you stop paying. A right to your data with no mechanism to retrieve it is not worth much.
• Closeout obligations. For subcontractors especially: what you are entitled to when a project ends, separate from when a subscription ends.
• AI training terms. Whether your data trains the vendor’s models, whether you can decline, and who owns improvements derived from it. This clause did not exist in most contracts two years ago.

If the answers are good, stay where you are. Switching construction software is expensive and disruptive, and “our vendor answered these questions well” is a perfectly good reason not to move. If they aren’t, and you do start looking, we have written up how WorkHoist compares to six named products — including where each of them beats us.

Where WorkHoist stands

WorkHoist sells software, so this section is the test of whether the rest of the page was worth reading. Here is where WorkHoist is genuinely strong on this, and where it is not.

MCP, on every plan

WorkHoist exposes the system over MCP, an open protocol, on every plan. That means your data is readable and writable from outside the product by an AI client you choose — Claude, ChatGPT or anything else that speaks MCP — rather than only through WorkHoist’s own interface. It is not an export you request; it is standing programmatic access to your live data, and it is not gated behind a higher tier.

This is a structural answer to lock-in rather than a promise about one. A system you can read and write from outside is one you can leave.

Bob, the in-app assistant, is a different thing

Bob is WorkHoist’s built-in assistant and it is read-only: Bob can answer questions about your jobs, and Bob cannot change anything. The write access described above comes from MCP, not from Bob. These get blurred together in marketing copy, including ours, and they are not the same capability.

Total Ownership

WorkHoist also offers Total Ownership, where you run the code on your own infrastructure. It is a bespoke engagement agreed per deal rather than a plan you can buy from the pricing page, and the terms are settled in that conversation.

Where WorkHoist is weak

Three things, stated here rather than left for you to find.

• WorkHoist is a younger product with a shorter integration list than the established platforms. If your data needs to move between WorkHoist and something specific, check that the connection exists before you commit to it.
• Total Ownership is a conversation, not a button. It cannot be bought from the pricing page, and a route to your own infrastructure that requires negotiating a bespoke deal is a weaker guarantee than one that does not.
• The Bob assistant sends your data to a third party. That is the next point, and it belongs in the open.

WorkHoist’s own subprocessor

When you ask Bob a question, the relevant parts of your business data are sent to OpenAI to generate the answer. OpenAI is a subprocessor of WorkHoist, based in the United States. A guide about data ownership that omitted its own author’s subprocessor would fail its own checklist, so here are both halves of it.

MCP connections are separate again, and worth understanding before you enable one. They are off by default and only an administrator can turn them on. Once connected, your data goes to whichever AI provider you chose, under their privacy policy rather than WorkHoist’s. WorkHoist cannot control what that provider does with it. Connections can be revoked at any time in Settings, and revocation takes effect immediately.

Questions contractors ask

Who owns the data in construction management software?

In most construction software contracts the customer owns their own data and the vendor holds a licence to process it. The practical question is not ownership but access: whether the customer can export the data in a usable format, and whether they still can after the contract or the project ends. Ownership is a legal claim and access is an engineering fact, and a contractor can hold clear title to data in a system they can no longer log into.

Can I export my data from Procore?

Yes. Procore provides the Procore Extracts application and the Data Extracts 2.0 tool, both of which export a project to files and folders. Procore’s own support documentation notes three limits worth knowing: most items are downloaded as PDFs rather than structured data, Procore Extracts covers only projects set to Active, and a Data Extracts 2.0 extract is available for download for 30 days. Whether that is sufficient depends on whether you need documents you can read or records you can import.

What happens to my project data if I switch construction software?

What you can take with you depends on the export your current vendor offers and the format it produces. Structured exports such as CSV, JSON or an API can be imported into a new system; PDF exports generally cannot, and usually have to be re-keyed. Check the export before you commit to a migration date, and check whether export access survives cancellation, because some platforms deactivate an account immediately when payment stops.

Does construction software use my data to train AI?

It depends on the vendor, and it is now a live commercial question rather than boilerplate. Construction platforms are competing on whether customer project data can be used to train their AI models, because a platform with data from thousands of projects builds better models than one without. Ask the vendor directly whether your data trains their models, whether you can decline, and who owns improvements derived from it.

What should a construction software contract say about data?

Four things: that the customer owns their data, stated explicitly rather than implied; a defined exit path covering what can be exported, in what format, and for how long after cancellation; closeout obligations covering what a subcontractor keeps when a project ends rather than when a subscription ends; and AI training terms covering whether customer data trains the vendor’s models and who owns the resulting improvements.