What is shipping, what is in progress, what is next
We label features clearly. Available means it works today. In progress is being built. Planned is on our list. Enterprise is scoped per engagement.
Core platform
Document upload + 9-stage processing pipeline
PDF / DOCX / HTML / CSV / TXT, sectioning, entity extraction, embeddings, claim extraction. OCR fallback for scanned documents is in progress.
Hybrid search (lexical + semantic + metadata)
Reciprocal rank fusion across BM25, vector cosine, and metadata filters.
Evidence-backed Q&A with citation validation
Every answer cites the source paragraph; verification pass flags unsupported claims.
Cross-document contradiction detection
Designed to surface when two documents disagree on facts, figures, or commitments.
Custom extraction templates
Field-level types, validation rules, confidence thresholds. Run across batches.
Compliance and risk rules
Deterministic condition types: keyword presence and absence, regex match, numeric thresholds, and missing-disclosure checks. Configurable severity and weighting per rule.
Human review queues
Low-confidence extraction results and answers auto-route to reviewers; approve, reject, or edit; full decision history.
Document version comparison workflow
Side-by-side diff between document versions with clause-level change detection.
Saved investigations / workpapers
Persist a series of related questions, answers, and citations as a single auditable workpaper.
Evidence-table builder
Build a structured evidence table from selected citations across multiple Q&A sessions.
Document analytics dashboards
Sentiment, theme and metric-trend analysis across a corpus, presented as dashboards. Not yet built.
Regulatory framework mapping
Auto-map extracted disclosures to TCFD / GRI / SASB / ESRS / IFRS S1-S2 requirement sets with coverage reporting. Not yet built.
Exports and reporting
Dataset exports (CSV / XLSX / JSON)
Export extraction results in any of three formats. Audit-logged.
Report generation (HTML / DOCX / JSON)
Generate evidence-backed reports with extraction tables and rule results.
Evidence pack export for Q&A
Bundle a Q&A answer + citations + workspace context as a structured JSON pack. Available now from the Search page.
Evidence pack PDF format
Render the JSON evidence pack as a print-ready PDF for hand-off to legal/audit.
Integrations and automation
REST API with multi-tenant authentication
JWT or API key authentication. Webhooks with HMAC signatures.
Webhooks with retry + auto-disable
Document and workflow events fire to your endpoint with HMAC-SHA256 signatures and exponential retry.
SharePoint connector
Native ingestion of documents from a SharePoint site. Permission-aware.
Google Drive connector
Native ingestion from a Drive folder with delta sync.
Salesforce / DocuSign integrations
Push document outputs back into CRM / e-signature workflows.
Operations and admin
Workspace-based RBAC (28 granular permissions)
Owner / Admin / Member / Viewer roles, granular permission gates per endpoint.
Full audit log on sensitive actions
Uploads, AI generation calls, review decisions, rule overrides, configuration changes, with actor, target and before/after values.
API key management
Hashed, prefix-indexed, optional expiry, last-used tracking.
SSO via Google OAuth + generic OIDC
Enterprise sign-on without password management overhead.
Admin retention controls
Per-resource retention windows configurable per workspace by an Admin.
SAML 2.0 SSO
Native SAML support for IdP-initiated SSO.
Compliance and security
GDPR Article 15 export + Article 17 erasure workflows
Self-serve from the user settings page. Bulk export via API on Business and above.
Encryption at rest (AES-256) and in transit (TLS 1.2+)
Standard infrastructure-level encryption.
DPA template + SCCs
Standard Data Processing Addendum and EU SCCs available during procurement.
Incident response plan documented
Severity classification, communication SLAs, post-incident reviews.
SOC 2 Type II audit
SOC 2-aligned controls in place. Audit timeline shared during procurement.
Data residency (EU + additional regions)
EU residency available now on Enterprise. Additional regions on request.
Dedicated single-tenant deployment
Single-tenant deployment in your VPC of choice on Enterprise engagements.
AI provider options
Managed AI by default with per-org cost caps
DataStruct AI runs on managed AI infrastructure with hard daily and monthly cost caps per organisation. Every AI provider call (generation, reranking, and embeddings) is audit-logged.
Customer-managed AI provider keys (BYOK)
Bring-your-own OpenAI / Anthropic keys so customers control provider billing, account-level retention, and procurement. Targeted at Business and Enterprise plans.
Azure OpenAI routing
Route AI calls through a customer-controlled Azure OpenAI deployment for procurement-driven environments.
Provider retention controls (Zero Data Retention)
Configurable provider retention modes including Zero Data Retention where the customer's provider account is eligible.
Redaction-before-LLM mode
Optional redaction of common PII (emails, phone numbers, identifiers) before evidence chunks are sent to the AI provider.
AI usage dashboard for customers
Customer-facing dashboard showing today and month-to-date AI usage against contract limits.
Private / self-hosted model routing
Route AI calls through a customer-hosted model endpoint for fully air-gapped enterprise deployments.
Internal evaluation
Internal accuracy benchmark dashboard
Track extraction accuracy, citation validity, abstention rate, and contradiction precision/recall against a labelled benchmark set. See docs/EVALUATION_PLAN.md.
Have a feature you need?
We move planned items up the list when an Enterprise customer needs them. Tell us what would unlock a deal.