Does Capveon replace our CRM or ERP?

No. It reads them and, where you enable it, writes to them after approval. You keep the systems and the contracts you already have.

What data do you need to start?

Read access to your CRM and a list of past projects or accounts. Nothing else is required for the first version.

Where does the public data come from?

Public federal, state, county and municipal sources: procurement portals, agenda platforms, capital plans, revolving fund lists, design awards and specification books. Every record cites its source.

Which AI models do you use, and what do they see?

Capveon uses commercial and open models under enterprise agreements that prohibit training on our inputs. Models see the documents needed to answer a specific question, not your whole tenant.

Will you complete our security questionnaire?

Yes. We answer questionnaires and share our policies under NDA. Write to hello@capveon.ai and a founder replies.

Your data stays yours. A person approves every write.

Capveon connects to systems you already run, so the questions are fair: what can it read, what can it change, and where does the data go. These are the answers.

Book My Demo

Grand Coulee Dam spillway, Washington. CC BY 2.0, David Brodbeck.

Commitments

Each one is enforced in the product, not in a policy document.

Read-only by default: Every connection to a system you run starts read-only. Writeback is enabled per system and per workflow, by you, and can be revoked the same way.
Human approval before writeback: No record is created, updated or sent without a person on your team approving it from a queue that shows the evidence.
Every claim linked to a source: Each field on a record opens to the document and page it came from. Where the record is silent, the field says unknown.
Per-field audit trail: Who approved what, when, and from which documents. Kept for the life of the record and exportable.
Tenant isolation: Nothing you connect is used to improve another customer's results.
Your data is not training data: We do not train models on customer data, and our model providers are contractually barred from doing so.
United States residency: Customer data is resident in the United States.
Single sign-on: SSO and SAML against your identity provider. Provisioning follows your directory, so someone who leaves on Friday has no access on Monday.
Access rules apply to AI answers too: A document a user cannot open is not summarised for them.
Deletion: Disconnect a system and its data is removed from your model. Leave Capveon and the tenant is deleted, with written confirmation.

A CRM your team trusts is worth more than the seconds an approval costs. The approval stays.

Questions security teams ask

Does Capveon replace our CRM or ERP?: No. It reads them and, where you enable it, writes to them after approval. You keep the systems and the contracts you already have.
What data do you need to start?: Read access to your CRM and a list of past projects or accounts. Nothing else is required for the first version.
Where does the public data come from?: Public federal, state, county and municipal sources: procurement portals, agenda platforms, capital plans, revolving fund lists, design awards and specification books. Every record cites its source.
Which AI models do you use, and what do they see?: Capveon uses commercial and open models under enterprise agreements that prohibit training on our inputs. Models see the documents needed to answer a specific question, not your whole tenant.
Will you complete our security questionnaire?: Yes. We answer questionnaires and share our policies under NDA. Write to hello@capveon.ai and a founder replies.

Trinity Lake behind Trinity Dam, California

The first workflow takes an afternoon, not a quarter.

Tell us the systems you run and the process that wastes the most time. We connect to them read-only and build the first version with your team in an hour or two, without replacing anything.

Book My DemoThirty minutes. Nothing to install.