Most data governance programmes die in the steering committee. Someone writes a 60-page policy, buys a catalogue tool, runs three meetings, and by the fourth nobody turns up. Nothing changes until a privacy question, an audit or an AI project forces the issue.
Then every question arrives at once. Where do we hold customer personal information? Which systems feed this board report? Who gave that vendor access? Which of our tools make decisions about people? Without basic governance, each answer is a week of someone digging through systems.
What follows is a data governance framework Australia's growing businesses can actually keep running: four steps, four artefacts, and named people instead of committees. It supports privacy compliance, and it makes AI work faster and safer.
Note: This article is general information, not legal advice. Check the official guidance and get advice for your situation.
The Privacy Act deadline that made this urgent
The Privacy and Other Legislation Amendment Act 2024 (No. 128, 2024) amended the Privacy Act 1988 across a long list of areas, including security, retention and destruction, eligible data breaches and penalties. Part 15 of Schedule 1 is titled "Automated decisions and privacy policies".
That part has a firm start date. As Allens summarised on 2 October 2026, the new disclosure obligations commence on 10 December 2026. Where a computer program makes, or does something substantially and directly related to making, a decision that could reasonably be expected to significantly affect an individual's rights or interests, the privacy policy must describe the kinds of personal information used and the kinds of decisions involved.
The OAIC released its final guidance on 30 September 2026. According to Allens' summary, it reads "computer program" broadly: rule-based tools and everyday software, spreadsheets included, can be caught, not only AI. Our post on what your software must disclose about automated decisions covers the obligation in detail.
Here's the practical problem. You can't describe what you haven't mapped. A lightweight governance framework is the cheapest way to know what data you hold, where it flows, who can reach it, and which systems use it to make decisions about people.
A data governance framework Australia teams can start with
Each step produces one simple artefact. Between them they answer most of what regulators, auditors, customers and AI projects will ask.
| Step | Question it answers | Artefact | Who owns it |
|---|---|---|---|
| 1. Ownership | Who's accountable for this data? | Data owner register | Business leads, one per domain |
| 2. Lineage | Where does it come from and where does it go? | Lineage map and data catalogue | Data or engineering lead |
| 3. Access | Who can see and change it, and why? | Access matrix with sensitivity tags | System owners, approved by data owners |
| 4. Quality | Can we trust it? | Quality rules and a scorecard | Data owners, with engineering support |
Step 1: Give every domain a business owner
Split your data into a handful of domains: customers, products, finance, people, operations. Give each a named owner from the business, not from IT. The owner decides definitions, approves access and signs off quality thresholds.
The register fits on one page: name, domain, systems of record, key terms, and a deputy for when they're on leave. If a domain has no willing owner, write that down. It's your first finding, and usually the most useful one.
Step 2: Trace lineage, start the catalogue
Data lineage shows how data moves from source systems through pipelines into reports, models and external parties. Don't try to map everything. Start with your ten most important reports and any system that makes or informs decisions about people, and trace each back to its sources.
Record what you find in a plain data catalogue: dataset, description, owner, source, refresh frequency, whether it holds personal information, and what reads from it. A spreadsheet is fine for the first version. If your pipelines are code, OpenLineage, "an open framework for data lineage collection and analysis", tracks datasets, jobs and runs, so integrations for tools such as Airflow, Spark and dbt can emit lineage events as jobs execute rather than someone drawing diagrams by hand.
Lineage is also how you find automated decisions you didn't know you had. A pipeline feeding a scoring rule, a triage queue or an eligibility check is exactly what the new privacy policy disclosures are about. Flag each one for privacy review.
Step 3: Control access with tags, not tickets
Tag personal and sensitive fields in the catalogue: contact details, government identifiers, health and financial information, anything about children. Then write an access matrix that maps roles, not individuals, to what each can read and change.
Enforce it in the platform: role-based access, and column masking for people who don't need to see raw values. In Snowflake that's a masking policy attached to a tag; in Microsoft Fabric or Azure SQL it's dynamic data masking; on AWS, Lake Formation handles column-level permissions. Review the matrix quarterly with the data owners, and include vendors, integrations and AI tools. Anything that receives personal information belongs on the list.
Step 4: Agree quality rules and publish a scorecard
For each critical dataset, agree three to five rules with the owner. Completeness, uniqueness, valid values and freshness cover most needs. Automate them in the pipeline and publish a one-page scorecard each month.
This matters for privacy as much as for reporting. Inaccurate personal information leads to wrong decisions about people, and records kept long past their purpose add breach exposure without adding any value.
Rules that keep it from becoming paperwork
- Govern what's used. Start with the data behind decisions, reports and AI. Ignore the rest for now.
- Put governance in the pipeline. Tags, tests and lineage that live in code stay current. Word documents drift within a quarter.
- One 30-minute meeting a month. Data owners review the scorecard, new access requests and new uses of data. That's the whole committee.
- Hook it to change. New system, new vendor or new AI feature? Update the catalogue and access matrix before go-live, as a release checklist item.
- Track two numbers. The share of critical datasets with a named owner, and the share passing their quality rules.
How JRC Software helps
Our data engineering, analytics and insights team sets governance up as part of the platform rather than as a separate programme. Week one is short workshops: we agree the domains, find owners, and pick the ten reports and the decision-making systems to trace first.
By the end of the first phase you have an owner register, a starter catalogue with lineage for those critical flows, sensitivity tags and an access matrix. You also get an inventory of systems that use personal information to make or support decisions about people. Your privacy and legal advisers can work from that when they update the privacy policy. We don't write the policy wording; that's their call.
We then put the controls into the pipelines: tags, role-based access, masking, automated quality tests and lineage capture. An enterprise catalogue can wait until you have owners; buying one first is money spent on empty shelves. A tool without owners is just a more expensive spreadsheet.
If you want the basics in place before 10 December 2026, book a governance consult with our team.
Frequently asked questions
What is a data governance framework?
A data governance framework is the set of roles, rules and tools that decides who's accountable for data, how it may be used and how its quality is kept up. For a growing business it can be small: named data owners, a catalogue with lineage, an access matrix and automated quality rules. The aim is trusted, well-managed data, not bureaucracy.
What is the difference between data lineage and a data catalogue?
A data catalogue is an inventory of your datasets, with descriptions, owners, sensitivity tags and refresh schedules. Data lineage shows how data flows between them, from source systems through pipelines to reports, models and external parties. The catalogue tells you what you have. Lineage tells you where it came from and what breaks if it changes.
Does data governance help with Privacy Act compliance?
Yes, a lot, though it doesn't replace legal advice. Knowing where personal information is held, who can access it and which systems use it to make decisions about people is the groundwork for privacy policies, security controls and breach response. It also makes it much easier to describe automated decision-making accurately once the new transparency obligations commence on 10 December 2026.
How long does it take to set up basic data governance?
For a mid-sized business, a first version of owners, a catalogue of critical datasets, an access matrix and core quality rules usually takes six to ten weeks. Coverage then grows as new systems and use cases come along. Starting with the data behind key decisions keeps the effort focused and shows value early, which is what keeps people turning up to the monthly review.
Australian Compliance
Book a governance consult
Tell us where you're stuck and we'll come back with practical next steps.
Get in touchSources
- OAIC: New resources on transparency for use of AI and automated decision-making
- Federal Register of Legislation: Privacy and Other Legislation Amendment Act 2024
- Allens: Mandatory automated decision-making disclosure requirements are coming, key takeaways from the OAIC guidance
- OpenLineage: An open framework for data lineage collection and analysis