Where your pricing, cost, and project data goes, and where it does not
Your estimator pasted the bid tab into ChatGPT. He was trying to save time. He did not tell anyone.
Revelry keeps your project data inside your control.
No credit card required.
Where the exposure actually is
The question is not whether AI touches your pricing.
Ensure complete visibility and control over how AI handles your pricing data.
Your people are already using it. The estimator who dropped a scope sheet in to summarize it. The PM who pasted an owner's contract clause and asked what it meant. The coordinator who ran a sub's quote through to pull out the exclusions. Nobody was being reckless. Every one of them was doing the job faster with the tool sitting in their browser.
What went with it: your unit costs, your labor rates, your markup structure, sub pricing you are contractually obligated to keep confidential, and in some cases owner documents covered by a confidentiality clause in the prime contract. There is no log of any of it. If an owner asked you next quarter to attest to how their project data has been handled, you could not answer the question.
That exposure is the real one. It is not theoretical, it does not require a breach, and it is happening in most GC offices right now.
The six commitments below are what we hold to instead.
The commitments
What we hold to instead.
SOC 2 Type 1 report
Independently audited controls over security. Report available under NDA. Type 2 audit in progress.
No training on your data
Revelry Ai does not use your project, pricing, and cost data to train models. Model providers work under API agreements that generally prohibit them from training on API data. We do not train or host our own models, and we do not opt into any provider program that would use your inputs for training. Model providers may keep a record of requests for compliance and abuse monitoring. Enterprise customers who bring their own model provider account can arrange zero data retention directly with that provider.
Your systems stay the record
Data stays in Procore, in BuildingConnected, and in your files. Revelry reads where you point it, and it does not become the place your records live. Anything that writes back to a connected system is held for a person to approve, and the decision is logged.
Access is permission-scoped
People see what their role allows them to see. Security events are audit-logged. Nothing gets read by someone your permissions did not already let read.
Named model providers, no surprises
Revelry runs on commercial models from established providers. We will name every subprocessor in the stack, and we will tell you before that list changes.
Every answer is cited
Answers carry citations to the source document and system. Your team checks the work rather than trusting it.
The answers your IT lead will ask for
Where it runs, and how long we keep things.
- Tenancy.
- Depending on the deal, you run on a white-labeled install of your own or in shared infrastructure with logical separation between tenants. Which one you get is written into the contract rather than assigned by default.
- Where it runs.
- The platform is built to be cloud agnostic. In practice it runs on AWS, and AWS is the subprocessor named in the SOC 2 report.
- Retention.
- Encrypted database backups are held for thirty days. Your documents and query history stay for as long as we need them to provide the service.
- Model providers.
- On a white-label install you choose which model providers process your data. On shared infrastructure you get the named list and notice before it changes.
Bring your IT lead in first.
Let them work through the security questions before your team uploads any project data.
No credit card required.