TellParrot, explained for machines.

Capabilities · current and shipped

Data Governance (pillar 01) /pillars/data-governance

• Roles and a governance council mapped to real systems and records, so 'whose call is this?' has an answer
• Policies and procedures attached to the systems, domains and records they govern, with coverage reporting and governance incidents
• Governance framework and maturity assessments aligned to DAMA-DMBOK and DCAM, with remediation and uplift objectives
• Assessment studio: bring your own assessment blueprints, run them across many respondents, and turn results into frameworks, maturity scores and remediation plans
• Adaptive conversational assessments: an interview mode with longitudinal memory, pre-scan of the estate, triangulation across respondents, and a governed commit of what was learned
• Brains: an organisation model with a memory. A brain for every unit, role, person, system and the company; briefs are assembled deterministically from counts with their sources (no language model writes them); a knowledge gate, witness records, belief states and vouching before knowledge is accepted; a scenario studio to rehearse a reorganisation. An advanced module an admin enables per tenant
• Data domains, departments and people modelled and linked to the systems and golden records they own
• Retention, privacy and compliance: retention enforcement, data subject access requests, erasure-versus-retention impact analysis, data journey and purpose
• Governance rules, tasks, an approval queue and automatic tasks, so governance is a worklist rather than a document

AI Governance (pillar 02) /pillars/ai-governance

• A complete register: AI initiatives, projects, proposals, models and agents, each with an owner and a system classification
• Risks and controls mapped to the EU AI Act and NIST AI RMF, impact assessments, an incident register, public disclosures, and contestability and appeals workflows
• Every model carries a weighted data-compliance score from the governance state of the data it reads, and a data-retention compliance view
• Agents under approval: AI agents inside the platform act only through proposals a person approves, and the ledger records who approved what
• Portfolio-level posture: an evidence and compliance register, the documents you would hand an auditor
• An AI readiness view that shows the thirteen layers under the model, from source systems to tribal knowledge, and which pillar owns each

Data Catalog (pillar 03) /pillars/data-catalog

• Smart ingest and collectors discover and profile assets as they arrive: column statistics, null rates, cardinality and distribution patterns, kept current continuously
• Deterministic classification that finds personal and regulated information: the same input always classifies the same way
• Column-level lineage from source columns through transformations to golden records and reports, with impact analysis before a change
• Asset profiles, an asset lineage graph, catalog applications, an ecosystem map and metadata layers, plus a data dictionary
• Trust scores per asset built from quality, freshness and governance state
• A business glossary whose terms are live definitions over real records, each with a live count of what currently qualifies
• Documents included: SharePoint and unstructured documents synced, scanned for personal information, and linked by mention onto golden records, with a subject-access view
• Microsoft Purview Data Map read, Power BI assets in the catalog, and publishing to Unified Catalog

Data Quality (pillar 04) /pillars/data-quality

• Six-dimension quality scoring per record, source and system, with trends, monitors and dimension breakdowns
• Quality gates on the way in: a preview of what processing will do before you commit, and held-for-review records that never reach a report, a customer or a model
• Quarantine and remediation, repair rather than rejection: transform and record fixes, CSV repair, and an ingest diagnostics engine that explains mapping faults
• Nightly scoring that remembers when a record started failing, so quality has a history rather than a snapshot
• Drift and watchdog alerts, and AI agents that queue the worst-quality records for steward review
• Quality where you work: the quality field on the Quality page and per-record scores on the golden record itself

Master Data Management (pillar 05) /pillars/mdm

• Ingest from anywhere: file upload, a REST ingestion API with scoped machine keys, near-real-time inbound webhooks, Microsoft Dataverse and Dynamics 365, Business Central, Azure Data Factory, database and warehouse connectors (PostgreSQL, Azure SQL, BigQuery, Redshift), and SaaS connectors for fundraising and events (DonorPerfect, Raisely, Eventbrite), church management (Planning Center) and e-commerce (WooCommerce)
• Migration from SQL Server Master Data Services, with hands-off pipelines and standing connector sync schedules
• AI-assisted column detection and mapping with transform rules applied automatically on ingest, and an onboarding assistant that proposes each step and applies it only on approval
• Matching that scales: configurable per-field rules (exact, fuzzy, phonetic, token), blocking for large datasets, scored duplicate clusters, survivorship previews, mass review, and background matching runs
• Humans hold the pen: stewards approve or reject every merge, merges are never silent, and the ledger records the decision
• Golden records that explain themselves: field-level provenance to every source value, identifiers, hierarchies, relationships, a per-record explain log and a plain-language narrative
• Field-scoped source trust weights decide survivorship where no rule applies, and every weight change is ledgered with a required reason
• Golden-record rules with a run inspector that shows exactly which records each run changed, before and after
• Truly reversible: unmerge is a true inverse of merge that restores the original records from split receipts rather than re-creating them
• Time-aware: temporal topology and an estate timeline replay any record, or the whole estate, as it stood on any date, along system time or business-event time
• Any-order linking: records link by identifiers and edges rather than foreign-key joins, placeholder records hold the place until the source arrives, and link coverage can be re-run
• Narrative ingest: unstructured interactions and notes become extracted facts on golden records through a review queue
• Australian business register enrichment with checksum validation of ABNs

Single Source of Truth (pillar 06) /pillars/ssot

• Governed feeds everywhere: a full read and write API, near-real-time outbound webhooks, an MCP server for AI assistants, and a Power Automate and Power Apps connector, all drawing from the same governed source
• Native to your warehouse: golden records mirror into Microsoft Fabric as open Delta tables and stream to Azure Event Hub; export to Dataverse
• Exports with receipts: data leaves through a governed gate with export control, lineage on every export and delivery receipts; outbound webhooks are host-validated and signed
• Export builder and export streams: map the governed vocabulary outward to a target template, with value translations, and merge survivors name the identifiers they retired so retired records stop leaving
• An advisor that cites its work: plain-language questions answered from live records with sources shown, and proposed fixes that wait for approval
• Segments that stay true: business definitions live as governed segments with live counts, and campaign lists come back consent-safe with evidence attached
• Answers: a deterministic cross-source insight engine that surfaces what the estate says when systems are compared
• One operations view: every load, matching run, sync, export and scheduled job in a single jobs monitor

Lens · Audit & Evidence (pillar 07) /pillars/lens

• Detect, decide, act, prove: rule violations paint the canvas, an action-first panel proposes the fix, a person approves it, and the evidence chain records it
• A chain that proves itself: every action is hash-linked to the one before it, append-only and key-versioned; any alteration breaks verification visibly
• The action ledger: who proposed, who approved, in what role, on whose behalf, and refusals are recorded too
• Two-person control over the exact content being approved, with cases that collect the full decision trail and a decide queue at the top of the case list
• Evidence bundles a third party can verify independently with an offline verifier, and optional external anchoring of chain checkpoints with anchoring status shown in the product
• Time travel, built in: the state of any record or the whole estate on any date, including business-event time
• An investigation canvas: saved multi-record graph workspaces with a geo view, a timeline drawer, expand-by-edge, pinning, and attachment to a case so the investigation travels with the decision
• Trace mode: pick two nodes and Lens walks the estate graph between them, every hop shown with the evidence behind it
• An action graph: every executed action is a node showing what it changed, what it was approved against and what it set off next
• Delivery you can prove: governed exports carry receipts for the disclosure, the subject-access response or the regulator's bundle

Platform experience /platform

• My Estate: the home page opens with an estate pulse and a feed of what needs attention, and the Data Command Center shows the estate itself
• Living dashboards: dot-per-record visuals across sources, resolution, quality and the Lens decide orbit, each expandable to a full page
• Build and consume work modes across the seven pillars, lens-based navigation, and enterprise search across governance entities
• An in-app documentation portal that also grounds the help assistant, an internal QA and feedback board, and a copilot that honours each tenant's AI configuration
• Workflows and automation, tasks, an approval queue, stakeholder engagement loops and system data reviews

AI inside the product /pillars/ssot

• Bring your own AI provider keys, scoped per tenant: Azure OpenAI, OpenAI, Claude or Gemini; no tenant data leaves through a shared key
• Every AI proposal waits for a human decision: the onboarding assistant, MDM agents, the governance advisor and the copilot all work propose-then-approve
• The advisor is grounded on the tenant's own data and cites its sources; in-app help is grounded on the product documentation
• MCP tools expose governed records and Brains to external AI assistants under the same permissions as a user
• Brains briefs are deterministic string assembly from counted facts, not model output, so what the organisation says about itself is reproducible

ParityIQ · environment parity /parityiq

• ParityIQ compares platform configuration between environments (build, test, UAT, production) table by table, row by row, value by value, from read-only access or file extracts
• Three modes: build validation before UAT, a go-live gate through the cutover window, and ongoing assurance after every change window
• Findings are triaged once and accepted with a reason; every later run shows only what changed. Reports land in storage the customer controls and no copy is retained
• TellParrot proves the data half of go-live, ParityIQ proves the platform half; together they make one go-live evidence pack
• Roadmap: ParityIQ runs and triage decisions inside the TellParrot workspace on the evidence chain. Maturity is labelled exactly

TellParrot for Box /box

• A Box connector: a Box Platform app with server authentication (client credentials grant), authorised by the customer's Box admin, syncs a chosen folder and its subfolders on demand or hourly, daily or weekly
• Files are extracted (Word, PDF, Excel, text), scanned for personal data deterministically with no AI model, and linked to golden records through a steward review queue with an optional auto-approve threshold
• Files whose content has not changed are not downloaded again; files removed from Box leave TellParrot with their links
• Optional write-back: approved record IDs, detected personal-data types and a sensitivity value are written to each file's Box metadata (global.properties), leaving other metadata on the file untouched; personal-data values never leave TellParrot
• Roadmap: record events that start Box retention clocks, Box AI as an optional extractor for scanned documents, and governed agent access to Box content. Maturity is labelled exactly

Microsoft /microsoft

• Runs on Microsoft Azure in the region you choose, including Australian data residency in Sydney; a transactable Azure Marketplace offer for procurement
• Microsoft Entra ID single sign-on and SCIM provisioning; joiners, movers and leavers flow from the directory you already operate
• Microsoft Purview: read the Data Map, write governance verdicts, classifications, glossary and stewardship back through an approval rail, publish to Unified Catalog
• TellParrot for Fabric: OneLake Delta export and mirroring, Purview write-back through approvals, Dataverse export, Event Hubs streaming, and a Microsoft Governance Hub with live Entra ID and Purview discovery, available today
• Rolling out: live discovery for Fabric, Power BI, Power Platform and Teams, and per-tenant consumption reporting. Roadmap: processing push-down onto your own capacity. Maturity is labelled exactly
• Power BI connector, Power Automate and Power Apps connector, Teams and Outlook, SharePoint documents, Azure Data Factory ingest, Azure Event Hub streaming, Azure OpenAI as an AI provider

Security, deployment and operations /security

• Single-tenant environments; multi-tenant control plane with an optional dedicated database per tenant, or bring your own PostgreSQL
• Deployment on Microsoft Azure first, with dedicated environments on AWS and Google Cloud in the region you choose, rolled out in waves inside your change control
• Encryption in transit and at rest; tenant isolation; scoped permissions with explainable access decisions; roles, access grants and access enforcement per tenant
• API keys inherit their service account's role and cannot write the control plane; outbound connector and webhook egress is host-validated and signed; rate limits on the advisor and the API
• A tamper-evident platform audit trail, source observability, ingest health per connection, OpenTelemetry tracing available opt-in
• Usage analytics, email delivery status, and a settings hub for configuration and SSO