SoftDocket

Platform · AI

AI that explains itself — and never trains on your data.

A copilot, an anomaly engine, and a forecaster sitting inside every SoftDocket module. Opt-in, audit-logged, DPDP-aligned.

Capabilities

Four AI surfaces. One trust model.

Each surface is opt-in per workspace, scoped by role, and emits a full audit trail.

  • Plain-English copilot

    Ask a question in natural language. Get a structured answer with citations back to the source records.

    • RAG over your live data
    • Cited every claim
    • Refuses unsafe prompts
  • Anomaly detection

    Continuously scans visitor flow, billing, helpdesk SLAs, and parking occupancy for outliers.

    • Per-site baselines
    • Tunable sensitivity
    • False-positive feedback loop
  • Forecasts

    30/60/90-day forward views for receivables, ticket load, and visitor surges.

    • Seasonality-aware
    • Confidence bands
    • Drilldown to drivers
  • Auto-summaries

    Daily, weekly, monthly digest of what changed — written for your role, not the dashboard.

    • Role-aware
    • Owner-tagged actions
    • WhatsApp + email delivery

Architecture

How the AI layer sits in the platform.

Your data stays in your tenant. The model talks to a retrieval gateway — not to your raw tables.

  1. Surface

    Copilot UI, anomaly inbox, forecast widgets — embedded in every module.

    • Copilot
    • Anomaly Inbox
    • Forecast Widgets
  2. Orchestration

    Prompt routing, citation enforcement, refusal policy, cost guard.

    • Router
    • Guardrails
    • Cost guard
  3. Retrieval

    Per-tenant vector index + structured query layer; RBAC enforced at retrieval time.

    • Vector index
    • SQL gateway
    • RBAC filter
  4. Data

    Your live SoftDocket tables. Read-only access. Encrypted at rest + in transit.

    • Postgres
    • Object store
    • Audit log

Models never train on your data. Provider is configurable per workspace.

Why ops leaders pick it

Less dashboard-hunting. More answered questions.

  • From dashboard to answer

    Skip the chart safari — ask the question, get the number, get the citation.

    Avg 9s to answer

  • Catch the silent regressions

    Anomaly detection flags the receivables dip three days before the month-end report would.

    3-day early warning

  • Trust by default

    Citations on every answer, refusal on unsafe prompts, full audit log per query.

    100% audit coverage

Common questions

Trust, cost, and control.

Ready to try it?

Watch the copilot answer your real question.

30-minute live demo on a sandbox tenant. We'll seed it with your data shape, you bring the questions.

AI-Powered Platform — SoftDocket