Data readiness audit & roadmap
Source profiling, quality baselining, PII scan and an object and dependency inventory, turned into a scorecard and a prioritised backlog.
CAPABILITIES
Four pillars, delivered as one continuous journey. Each has named people behind it, a repeatable framework, and at least one delivered proof point, and each can be bought on its own or as part of a programme.
THE DELIVERY CHAIN
Most firms sell one link and hand the rest on. The handover is where programmes lose the thread: the advice given in week one stops binding the build in week twenty. Each pillar below passes something specific to the next.
So what: you do not have to buy the chain. You do have to know which link your problem sits in, buying analytics when the break is in reconciliation is how data programmes spend a year proving nothing.
PILLAR 01
Know what you actually have before you spend against it. We profile the estate, price the gap, and give you a sequence, not a maturity model with no next action in it.
Source profiling, quality baselining, PII scan and an object and dependency inventory, turned into a scorecard and a prioritised backlog.
Configurable DQ rule frameworks in SQL your team owns, de-duplication and standardisation, per-layer load audit, daily failure and remediation reporting.
PII classification and masking, RBAC and row-level security, retention policy with automated and cascading disposal, ISM and Essential Eight aligned handling, plus the AI use register and risk tiering the same controls will have to cover once models start reading this data.
Future-state data models, source-to-target mapping, data dictionaries and business glossaries, documented lineage across ingestion, transformation and reporting.
Proof: MetLife’s PII classification framework and enterprise data dictionary; Uniting’s future-state models, which cut data-source onboarding time by 60%. Read the case studies →
THE READINESS AUDIT
You cannot price a migration or an AI programme against an estate nobody has profiled. The audit is deliberately cheap and deliberately uncomfortable, it is designed to find the things that would have surfaced in month four.
So what: the audit either tells you the programme is smaller than you feared, or it finds the issue that would have cost you the timeline. Both answers are worth two weeks.
PILLAR 02
Your cloud partner moves the infrastructure. We own the layer they don't: converting the business logic, and proving it survived the journey. Platform-neutral, the recommendation follows your workload, not a resale margin.
Which platform suits which workload, given what you already license. Snowflake, Databricks, Fabric, BigQuery, Dremio, PostgreSQL and MongoDB all on the table.
Object and dependency inventory, schema mapping, table / view / stored-procedure / job conversion, rehearsals, wave planning and the cutover evidence pack.
Medallion architecture, metadata-driven ingestion frameworks, and pipeline re-engineering off Informatica, SSIS, DataStage and hand-built ETL.
Terraform, environment promotion and CI/CD, plus workload analysis and right-sizing when the new platform costs more than the one you left.
Proof: Coal LSL: 100% of database objects converted, validated and deployed in three months; Transport for NSW. 35% monthly cloud cost reduction. Read the case studies →
PILLAR 03
AI projects rarely fail on the model. They fail on ungoverned, undocumented, unreconciled data underneath it. We build the foundation first, then one real use case on top of it, with a person signing off before anything is generated.
Quality baseline, metadata coverage, access controls and lineage gaps, an honest answer on whether a use case can succeed before you fund it.
Scattered content ingested and indexed for Snowflake Cortex Search or BigQuery with Gemini, with permissions respected at retrieval rather than bolted on after.
Agents scoped to one job with a bounded tool list, metadata and DQ agents, reconciliation and exception triage, document and claim intake, each with an approval gate and a logged trail of what it touched.
An AI register, risk tiering, accountable-owner mapping, disclosure and escalation paths, and the human-control points, written to the practices your regulator and your board are already asking about.
Golden-set evaluation and a regression record you can take to a board, grounding and refusal checks, prompt and tool-call tracing, drift and cost monitoring, and retrieval, caching and routing tuned so the per-query cost stays defensible.
Column-level descriptions and candidate DQ checks generated for newly onboarded tables, proposed by AI and approved by an engineer.
GOVERNED TO THE AUSTRALIAN GUIDANCE
The Commonwealth's Guidance for AI Adoption replaced the Voluntary AI Safety Standard in October 2025 and sets six essential practices. Most AI governance decks in this market still cite the old standard. We map your use cases to the current six, and leave behind the register, the test record and the escalation path, not a policy PDF.
PRACTICE 01 & 04
Decide who is accountable · share essential information
AI register with a named owner and a disclosure line per use case.
PRACTICE 02 & 03
Understand impacts · measure and manage risks
Risk tiering, a privacy and impact view, and the controls attached to each tier.
PRACTICE 05
Test and monitor
Golden-set evaluation, grounding checks and drift and cost monitoring in production.
PRACTICE 06
Maintain human control
Approval gates, bounded agent tool lists, and a kill switch that someone owns.
Aligned alongside the Privacy Act's automated-decision transparency obligations, the National AI Plan's sector-by-sector approach, and ISO/IEC 42001 where certification is the goal. APRA CPS 230 and 234, ASIC and DTA obligations handled where they apply.
Proof: The Smith Family: Zendesk and SharePoint unified in one governed search; Woolworths Group. LLM agents generating column metadata and DQ checks. Read the case studies →
THE RETRIEVAL DIAGNOSTIC
Enterprises that have already invested in data keep hitting the same wall: the AI assistant returns the wrong clause, the wrong version, or nothing. There are four root causes and they do not share a fix.
So what: the two right-hand quadrants are data work, and roughly half the cases we see land there. Replacing the assistant fixes the left column at best, which is why the diagnostic comes before the licence renewal.
PILLAR 04
One definition of the truth, in the hands of every business unit, and the feature layer underneath it reliable enough that data science is worth doing at all.
Conformed dimensions and metric definitions so finance, operations and marketing stop quoting three different numbers for the same week.
Power BI, Fabric, Looker and Tableau reporting on certified datasets, plus dashboards that surface DQ, disposal and reconciliation status to leadership.
Feature-ready datasets on governed data, segmentation, propensity, churn and demand models, evaluation design and MLOps hand-off to your platform team.
Identity resolution, consent and suppression, and audience and event pipelines from the warehouse into a CDP or Braze, reconciled back to the records that triggered them.
The metrics, definitions and documented contracts an agent or an assistant has to read before it can answer a business question, so natural-language analytics returns your numbers, not a plausible guess at them.
Proof: NSW Health, medallion reporting for the EDWARD programme; IAG. ACCC and APRA regulatory reporting logic. Read the case studies →
METRIC LINEAGE
Organisations rarely have a reporting problem. They have a definition problem that only shows up in reporting, and it gets worse the moment an AI agent starts answering from the same estate.
So what: an agent on an estate with no conformed model will answer confidently and differently each time, because it is reading whichever source it found first. The semantic layer is the prerequisite for natural-language analytics, not a nice-to-have after it.
THE BENCH
GCP · Snowflake · Azure certified
Fabric · Synapse · Databricks · Snowflake
Power BI · Fabric · SQL
Python · Databricks · Snowflake
Azure · AWS · GCP · Terraform
Onshore by design. Sydney-based, Australia-wide. Client IP, code and data stay in your Australian environment, suited to government, not-for-profit and regulated financial clients.
Every engagement ends with artefacts your team can run without us. That's the test we hold ourselves to, and the reason our handover doesn't come with a support contract attached.
DELIVERY
Two views we put in front of every prospective client and partner before scoping: how we compare against the other ways of buying this work, and how deep our competency actually runs in each domain. Both are deliberately unflattering in places.
HOW WE COMPARE
Solves the data problem, not a tool resale
LARGE SI
OFFSHORE VENDOR
INDIVIDUAL CONTRACTOR
THE DATA GUYZ
Client-owned, documented IP
LARGE SI
OFFSHORE VENDOR
INDIVIDUAL CONTRACTOR
THE DATA GUYZ
Depth across your existing multi-cloud stack
LARGE SI
OFFSHORE VENDOR
INDIVIDUAL CONTRACTOR
THE DATA GUYZ
Lean delivery overhead
LARGE SI
OFFSHORE VENDOR
INDIVIDUAL CONTRACTOR
THE DATA GUYZ
Onshore data residency & governance
LARGE SI
OFFSHORE VENDOR
INDIVIDUAL CONTRACTOR
THE DATA GUYZ
Team continuity & accountability
LARGE SI
OFFSHORE VENDOR
INDIVIDUAL CONTRACTOR
THE DATA GUYZ
Scale for 24x7 L1 support or 100+ FTE
LARGE SI
OFFSHORE VENDOR
INDIVIDUAL CONTRACTOR
THE DATA GUYZ
The last row is deliberate. We are designed to complement large SI and MSP scale, not to replace it. If you need 24x7 L1 support or a hundred-plus FTE bench, that is not us, and we will say so in the first conversation rather than the third.
Harvey ball: full = strong fit, half = partial, empty = weak fit. Analytical comparison for partner and client scoping; indicative.
Take 30 minutes with our principal architect across data, cloud, analytics and AI. No fee, no deck, we'll tell you which pillar actually applies, and in what order.