FAQ

Questions brokers ask us

Does ClaimNest decide whether a claim is covered?

No. ClaimNest prepares documentation and flags administrative issues. It never determines eligibility, approves or denies a claim, or presents an AI output as a binding decision. Those decisions belong to the insurer and its authorised staff.

How do I know where an extracted value came from?

Every extracted field and every inconsistency flag links to the source document and highlights the line it was read from. Reviewers verify values against the original before approving.

What happens if the AI reads a document wrongly?

Reviewers correct the value and record a reason. Corrections to critical fields (policy number, name, dates, amounts) proposed by non-reviewers need a reviewer's approval. Every change is recorded in the audit history.

Which documents are supported?

Claim forms, invoices, receipts, repair estimates, reports and photographs (using image metadata and vision-based descriptions). Amazon Textract handles supported PDF and image formats; unsupported or low-quality files are queued for manual handling and clearly labelled.

Can we change the checklists?

Yes. Checklists are configurable per claim type and per tenant, including which document types satisfy each item and whether it is required.

Where is our data stored and for how long?

Documents sit in encrypted, tenant-isolated Amazon S3 storage in your selected AWS region. Retention is configurable per tenant and claim type, with deletion jobs that are themselves audit-logged.

Is customer data used to train models?

No. Amazon Bedrock does not use inputs or outputs to train the underlying foundation models, and ClaimNest does not train on customer documents.

Is the demo on this site connected to AWS?

No. The interactive demo and the sample workspace run in your browser on fictional data to show the experience. The production architecture (S3, Textract, Bedrock, PostgreSQL, Lambda) is documented but not what powers the demo.

How does HubSpot fit in?

HubSpot manages our own sales process: demo requests, broker prospects, scheduling, pipeline and onboarding. It never receives claim data.