Certant: source-linked document AI for clients and internal teams

Build answers from the client's own documents

Certant gives consultancies, integrators and agencies a platform for delivering document-grounded AI under their own or a client's brand. Its described uses include policy chatbots, agreement checks and analytics built from an organization's reports.

The publisher says answers show the source page or paragraph, allowing staff to inspect the basis of a response. Agreement-checking agents compare documents with the client's rules, flag clauses and wait for a person to approve what happens next.

Choose the deployment and billing model

Certant distinguishes its credit-billed cloud service from deployments on customer hardware. The latter uses open-weight model files and a renewable fixed-term license without a per-question charge, according to the site. It also describes an air-gapped option and an ASUS Ascent GX10 appliance offering.

Those are deployment choices to examine during setup. A customer-hardware installation needs its model, dependencies, connectors and access controls configured consistently with the intended data boundary. The site lists single sign-on, per-document permissions and an audit log as platform features.

For partners, each client receives a separate workspace under an enterprise agreement. Implementation, integration and support remain services the partner can provide. Cloud credit billing and the customer-hardware license should be compared separately when estimating operating costs.

Check citations against difficult questions

Franklin analysis: start with a small collection of documents whose permissions and answers are already understood. Ask a question with a clear source, one involving conflicting versions and one whose answer is absent.

Inspect whether the cited paragraph supports the response and whether the system respects document access. For an agreement workflow, verify the flagged clause and the approval step before enabling any follow-on action. This evaluates the core proposition—useful answers tied to customer evidence—rather than treating a populated knowledge base as proof that every answer is grounded.