Black Shard

One data layer, and AI that does the work on it.

The data platform a business runs on, the reporting over it, and the AI automation that moves work through it, on Azure in Australian regions.

Talk to us about your data and automation

Most businesses hold their data across a line-of-business system, an accounting package, spreadsheets and a portal, and people carry it between them by hand.

AI automation, guardrails and client data

We bring it into one platform with a defined schema, build the pipelines that keep it current, and put reporting and dashboards over it. On that layer we build AI automation for the tasks people do by hand: document intake and extraction, classification and routing, drafting from records, transcription and speech. Every AI feature ships with guardrails, an audit trail and tenant isolation, the discipline we run in our own speech-to-text and document pipelines in production on Azure AI services in Australian regions. Client data is not used to train any model.

Data platform engineering

A single, governed data layer for the business: schema, storage, access control and history, on Azure in Australian regions.

What you get

  • A defined schema over the records the business runs on, with row-level access control where the data is shared
  • PostgreSQL and Azure storage sized and backed up for production, with point-in-time recovery
  • An audit trail of who changed what, kept for as long as the business needs it

Integration and pipelines

Scheduled and event-driven pipelines between the systems the business already runs, with every run logged.

What you get

  • Connectors to accounting, CRM, job and portal systems through their published APIs, including Xero as a listed developer partner
  • Incremental sync with retries, reconciliation counts and exceptions raised to a person
  • A run history that shows what was read, what was written and what was rejected

Reporting and dashboards

Reports and dashboards over the data layer, built for the questions the business actually asks.

What you get

  • Role-based dashboards for directors, finance and operations
  • Scheduled reports delivered to the people who act on them
  • Definitions documented once so every number reconciles to its source

AI workflow automation

AI built into the workflow at the step where a person re-keys, classifies or drafts, with a review queue before anything leaves.

What you get

  • Extraction, classification and routing of inbound documents and messages
  • Drafting from the record: letters, summaries and responses prepared for a person to approve
  • A review queue and approval step in front of every automated action that reaches a customer or a ledger

Document and speech intelligence

Optical character recognition, document understanding and speech-to-text on Azure AI services in Australian regions.

What you get

  • Document intake pipelines that read invoices, forms and correspondence into structured records
  • Speech-to-text and dictation capture, with the same evidence-and-audit discipline as our clinical platform
  • Australian endpoints, no training on client data, retention disabled where the service supports it

Guardrails and audit for AI features

The controls every AI feature carries before it reaches a customer: input and output guardrails, an audit trail and tenant isolation.

What you get

  • Guardrails enforced in code on what enters a prompt and what leaves it
  • An audit log of what the feature read, what it returned and what happened next
  • Tenant isolation at the data layer, tested with a cross-tenant read attempt before release

How it is shaped

A systems analysis first: every system, every data flow and every manual step mapped and costed. Then fixed-scope build phases, with a managed service after go-live on one monthly fee.

Recent work of this kind

  • Aurii

    Clinical software

    The spoken consult becomes a draft clinical note in the practice's fixed section order, and the specialist edits, approves and signs it.

    Read the case study

  • Restart Recruitment

    Recruitment

    An uploaded CV is parsed into a structured candidate record, returned as strict JSON against a fixed schema, and the original file is kept in Azure storage.

    Read the case study

  • Stone Leaf Capital

    Capital markets

    Document register in living sync with the firm's SharePoint estate, with provenance-guarded retention and a quarterly restore drill.

    Read the case study

The rest of the firm.

Tell us where the data sits and what the work is.

Brisbane head office.

Open a briefinfo@blackshard.com.au