Understand where your
proposal data goes.

From document to draft

1. Extract and review. The document upload page extracts text in your browser so you can review and edit it before submission.
2. Submit with your account. Submission requires a signed-in account with a verified email. The uploaded source is stored in Supabase Storage; reviewed text and proposal details are sent to Airtable for the proposal workflow.
3. Generate and deliver. Airtable AI processes submitted text to create and revise a draft. Completed drafts are sent through our email delivery service, Resend. Submitted text is processed by services beyond source-file storage.

Knowledge Base: uploads, indexing, and search

1. Store privately. Original documents are stored in a private Supabase Storage bucket. Extracted source text, citation references, reusable answers, and search embeddings are stored in Supabase PostgreSQL.
2. Consent to indexing. When you authorize indexing, the background processing service extracts text and sends source excerpts to the OpenAI API to create embeddings. Embeddings are numerical representations used for meaning-based search. The original file stays in source storage; its extracted text is processed by the embedding provider.
3. Choose your search. Meaning-based AI search sends your question to the OpenAI embedding API. Filename, source-text, answer-text, and activity text searches query the stored library without sending that search query to the embedding provider.
4. Review and reuse. People write reusable answers and owners or reviewers approve a specific saved revision. Editing the answer clears its approval. Source quotations and private original downloads support human verification.

Indexing requires your processing consent. Use only documents you are authorized to share with your organization and the processing providers.

Compliance analysis

The compliance workspace stores original RFP sources and extracted text in Supabase. When you review the extracted text and request analysis, bounded source excerpts are sent to the configured OpenAI API model to identify requirements and facts. Saved results include source quotations and human review fields.

Analysis does not prove that your company meets a requirement. Reviewers supply company evidence and references before marking a row compliant. Proposal submission is a separate step that sends the submitted inputs to the Airtable drafting workflow described above.

Access to source documents

RFP uploads and compliance analyses use account ownership checks. Knowledge Base access is scoped to active organization membership and role permissions; all active members can read the organization's sources and draft answers. Owners manage invitations and can deactivate members. Revoking access stops future requests but cannot recall copies someone already downloaded.

Storage buckets are private. Authorized operators and processing services may need access to submitted information to operate the service. Private storage does not mean that no service provider or authorized operator can access your data.

Retention, deletion, and AI data use

Our proposal workflow uses Airtable AI on the Team plan. Airtable's published AI terms prohibit using customer inputs and outputs to train generative AI models. On the Team plan, AI providers may retain inputs and outputs for up to 30 days for safety and compliance moderation, with exceptions for prohibited content and applicable legal obligations. Usage metadata has separate retention terms.

Knowledge Base embeddings and compliance analysis use the OpenAI API, separately from Airtable AI. OpenAI's applicable account agreement and API data controls govern those requests. Provider retention and data-use settings depend on the service and account configuration; we do not promise zero provider retention. Contact us to confirm the terms required for your organization.

AI-provider retention limits do not automatically delete proposal records, source files, extracted text, embeddings, or delivery emails.

Archive and deletion are different actions. Archiving a Knowledge Base source removes it from future retrieval and invalidates dependent answer approvals. Its original file, extracted text, embeddings, and audit history remain stored. Archiving does not request deletion from downstream providers.

Source files, reviewed text, generated drafts, and delivery messages can have different retention and deletion requirements. The document upload page does not currently provide self-service deletion. For a deletion request or questions about AI training and provider retention, contact our team. Please include your account email, but do not attach the confidential document.

Ask a security question or report a concern

Email hello@tendercrestai.com with the subject “Security inquiry” or “Security report.” Share a brief description and the affected page. Do not include passwords, access tokens, or customer document contents. We can arrange an appropriate way to exchange sensitive details.

Proposal project data

A project has explicit membership in addition to organization access. Sharing an RFP snapshot copies its private source files into the project; it does not transfer ownership of existing uploads. Saved revisions retain evidence references and human review history.

Generated submission packages are private, use short-lived authorized download links and expire after seven days. Project source copies and audit records remain until the project creator explicitly deletes the project. Knowledge Base originals and existing personal uploads follow their own retention process.

Workspace notifications are currently in-app. Project email reminders are not enabled for this release. The separate existing emailed proposal-draft service remains available. An approved revision represents a human review decision, not an independent security certification or a guarantee of compliance.

Optional AI writing in the proposal workspace

Whole-proposal assembly from accepted company answers, bid-fit calculations and rule-based quality checks run without an additional model request. When you explicitly choose AI regeneration, clarity, shortening or expansion and confirm processing consent, the saved section text, linked RFP requirements, company name and accepted company evidence excerpts are sent to the configured OpenAI chat API.

The result is saved as a private job preview. It does not replace your section until you accept it, and the changed revision needs human review. Writing requests use store:false; standard provider retention may still apply. This is not a verified Zero Data Retention arrangement.

Use only information you are authorized to process with these providers. Approved-answer reuse stays within your organization and does not initiate training of a shared TenderCrest model. Human reviewers remain responsible for factual claims, evidence scope and final submission instructions.

Public reference guidance

The optional project guidance feature matches saved requirement text to a versioned local collection of reviewed public references. It makes no additional model, embedding or external-source request. Public source links are provided for your inspection; opening one takes you to that publisher's website.

References guide response structure only. They are separate from private organization evidence, buyer requirements, compliance decisions and saved approvals. They are not automatically added to customer exports, used to train a shared model or updated from customer uploads.

Evaluate operational and contractual fit

Role-based access, private storage, revision checks and human sign-off are application controls. They do not establish independent certification. This site does not claim SOC 2 certification, SSO, a contractual uptime SLA or verified Zero Data Retention.

Standard provider retention may apply. Release operations currently accept Netlify's all-scope secret limitation; Functions-only secret scoping is unavailable on the current plan. Do not assume build-context isolation for server credentials.

Generated packages expire after seven days; this is not a seven-day deletion promise for all project or provider data. Recovery checks performed for a release are not a contractual recovery-time or recovery-point commitment.

Before sending regulated or confidential bid material, request a review of your data classification, provider terms, access requirements, deletion needs and recovery expectations. Ask for evidence of any certification or contractual assurance you require. Do not send confidential documents through the contact form.

Proposal history and deletion requests

Where account continuity is enabled, retained emailed-draft copies are private to the verified recipient account and expire after 90 days. Downloads reconstruct the Word draft from the retained text. Older email-only deliveries are not retroactively stored. Provider acceptance is distinct from delivered/bounced events, which require verified webhook configuration.

Personal upload deletion can remove an unused original only after explicit confirmation and dependency checks. Submitted, analyzed, shared or organization sources require a reviewed deletion request. A request is not completion; its status appears in My Account. Shared project copies, provider retention, financial/audit records, received emails and separately held backup copies are separate. Contact support to clarify the scope of a broader erasure request.

Optional scanned-PDF OCR runs locally in the browser, using pinned third-party engine/language downloads. The OCR page does not upload page images or the PDF. Recovered text is derived, unverified content; preserve the original and check all clauses, amounts and dates. Current limits are English, five pages and 10 MiB.