Support data controls that are clear enough to evaluate

Data flow

1. An email webhook or website widget maps a request to one organization.
2. Ticket messages and configuration are stored in tenant-scoped Postgres tables.
3. Before AI processing, AppsResolve replaces common email addresses, phone numbers, bearer tokens, secret assignments, private keys, and URL query values in customer-authored context. This is a practical reduction step, not a guarantee that every possible identifier is detected.
4. The redacted conversation, relevant approved documentation, workspace context, and bounded reviewed-ticket examples may be sent to the AI provider.
5. Each context item receives a stable source ID. Model-returned citations are accepted only when that ID was supplied and citation-eligible for the same run. Prior tickets are examples only.
6. The resulting summary, category, confidence, validated citations, source excerpts, version/freshness snapshots, conflicts, and draft return to the review queue.
7. A person opens the source, checks whether the context matches, edits, records a correction reason, asks for information, escalates, or explicitly approves.
8. An approved reply is sent through the configured support channel and remains in the ticket history.

The current drafting pipeline does not browse the web. External sources are shown as not searched rather than implied.

Vendors

Read the current integration boundaries.

AI processing and reviewer corrections

Anthropic receives only the selected, preprocessed ticket and knowledge context required for a processing request. AppsResolve does not train its own models on customer content. Anthropic states that commercial API content is not used to train its models unless a customer explicitly opts into its Development Partner Program. Anthropic publishes a standard API retention period of up to 30 days, subject to its documented agreements, policy-enforcement, and legal exceptions.

When a reviewer edits a draft, AppsResolve can record the reason, edit distance, draft-to-send time, cited-source keys, and whether a knowledge update should be reviewed. The edited reply does not become approved knowledge, a retrieval source, or training data automatically. A person must create, review, and publish any reusable knowledge.

Customers can define restricted topics and should avoid placing secrets, payment-card data, private keys, or unnecessary sensitive data in support messages or knowledge content. Managed-service reviewers can access in-scope tickets and approved documentation only to deliver the agreed service.

Access controls

Supabase Auth handles sessions. Organization access requires an explicit membership record. Tenant-scoped queries carry an organization context and Postgres row-level security policies provide a database boundary. Operator routes use a separate allowlisted identity. Access can be revoked by removing organization membership or operator authorization.

Managed support does not rely on invisible permanent workspace access. A support session requires a recorded reason and explicit scope. The default is read-only access to the organization summary. Sessions are limited to 15, 30, or 60 minutes, expire automatically, and can be ended early. The session start is recorded in the platform audit log, and an explicit early end creates a second audit event. Broader scopes or permitted changes must be deliberately selected.

Audit history

Sensitive platform actions are recorded with the actor, role, organization, target, action, and timestamp. This includes managed support access, organization administration, billing changes, pilot decisions, and reviewer assignments. Audit metadata shown in the operator console is restricted to an approved field list so ticket content and secrets are not exposed through the audit viewer.

Data retention, export, and deletion

Ticket and knowledge data is retained while an account is active so the support history remains usable. Automated retention schedules, self-service exports, self-service account deletion, and published backup-restore reports are still being developed.

Requests are handled manually through support@appsresolve.com after identity and organization authority are verified. The current deletion path archives the organization first, blocks permanent deletion while a Stripe subscription is active, removes tenant records in one database transaction, attempts attachment-object cleanup, and records the administrative result. Limited billing, security, legal, or deletion audit records may be retained when required.

Encryption

Production browser traffic uses HTTPS. Hosted providers document encryption for their managed services, but AppsResolve has not completed an independent encryption-control audit and does not claim a separate proprietary encryption layer.

Incident reporting

Report a suspected security or privacy issue to support@appsresolve.com. Include the affected organization, approximate time, and a description that avoids unnecessary sensitive data. ND SOFT will triage the report, restrict affected access when needed, preserve relevant audit evidence, investigate the scope, coordinate with involved providers, and communicate with affected customers when required. No fixed acknowledgement time or service-level guarantee is claimed yet.

Compliance status

AppsResolve does not currently claim SOC 2, ISO 27001, HIPAA, PCI DSS certification for the application, or another formal security certification. Stripe handles card payment data in its own environment. Security documentation and controls will be updated as the product and customer requirements mature.