News

What the OECD Digital Government Outlook 2026 means for AI in Australian government agencies

The report shows that data is becoming essential to how governments operate.

September 21, 2026

Data & Government AI working for Australians

Is government AI delivering results? How do we know if government AI will improve public services?

TL;DR

Is government AI improving public services? What the OECD Outlook means for Australia

The OECD Digital Government Outlook 2026 shows that data is becoming essential to how governments around the world operate, but many still struggle to turn national data strategies into everyday practice.



What the report tells us about government data:

  • Countries are improving how they manage data. Since 2023, the OECD’s Digital Government Index has increased by nine percentage points, while its open data index has risen by five percentage points.
  • Data sharing is still inconsistent. Governments are investing in cloud platforms and systems that allow agencies to exchange information, but only 63% of public institutions actively use national data-sharing systems.
  • Good strategies do not always lead to action. Many governments have strong digital and data strategies but find it harder to improve data quality, reuse information and apply those strategies across agencies.
  • Better services depend on connected data. Secure data sharing between agencies can help governments provide simpler, more joined-up services designed around people’s needs.
  • Trust remains a major challenge. Only 52% of citizens trust governments to use their personal information appropriately.

How do we know if government AI is improving public services?

Governments are rapidly adopting AI, but the OECD found that 75% of the countries surveyed do not evaluate whether digital and AI investments achieve their intended results. The next phase of government AI should be judged by whether it improves services, decisions and outcomes, not by how many strategies, platforms or pilots are launched.

A government agency can launch an artificial intelligence system on time and call the project a success. But unless it can show that the system made a service faster, fairer, easier to use or less expensive, what has actually been achieved?

The OECD’s Digital Government Outlook 2026 suggests many cannot answer that question. It found that 75% of the countries surveyed do not evaluate whether their investments in digital technologies and AI deliver their intended results.

At the same time, 92% have an AI-in-government strategy, and AI is now used in almost every OECD government, particularly to improve internal processes and public services.

Pierre du Preez, Director at Notitia, says what is often missing is evidence that strategies are producing better services, better decisions or better outcomes for the public.

“An AI strategy can show that an organisation intends to act, but it cannot show that the investment is working,” Pierre says.

“Before scaling AI, organisations need to be clear about the problem they are solving, whether the underlying data can support that use case and how success will be measured.”

A strategy is not evidence of success



The OECD’s report draws on its Digital Government Index and Open, Useful and Re-usable Data Index. It covers 36 OECD countries and eight accession countries. Australia is included in the research and has a dedicated country note.

The results show meaningful progress. Since 2023, the average Digital Government Index score has risen by nine percentage points, while the average open-data index score has increased by five percentage points. Every participating country now has a government data strategy, and 69% have made more than half of their online services accessible using a digital identity.

However, the OECD finds that governments are generally better at developing strategies than implementing or monitoring them.

Most countries have data-interoperability systems, for example, but only around two-thirds report widespread adoption across government. Digital identity systems are well established, yet they do not always make public services straightforward to access.

AI presents the same problem at greater speed. Governments are experimenting with AI and establishing strategies, but evaluation, procurement support, risk assessment and transparency have not advanced at the same rate.

In other words, governments have become better at setting a direction. Turning that direction into routine, measurable improvement remains much harder.

Australia already has the policy foundations

Australia’s Data and Digital Government Strategy sets a vision for simple, secure and connected public services delivered through data and digital capability.

The Australian Government has also introduced a policy for the responsible use of AI in government, supported by guidance for public servants using generative AI.

These are important foundations. They tell agencies where the government wants to go and establish expectations for responsible use. They cannot, on their own, show whether an individual investment has improved anything.

  • Did the technology reduce the time required to deliver a service?
  • Did it improve the quality or consistency of a decision?
  • Did it remove repetitive work without introducing new risks?
  • Can the organisation explain how the system reached an outcome?
  • Are the results experienced by the people who use the service, rather than only appearing in a project report?

Without questions like these, success can become defined by completing a pilot, deploying a platform or reporting how many people used a tool. Those measures tell us that something happened. They do not tell us whether it was worthwhile.

Where does data come in? What this means for Australian government agencies

For Australian government agencies, the challenge is to turn national policy into measurable improvements within individual services and programs. That requires more than selecting an AI platform or completing a pilot.

Each proposed use case needs a clearly defined service problem, an agreed starting point and evidence that can show whether the investment has improved the result. Agencies also need to determine whether the relevant data is accurate, accessible, appropriately governed and suitable for the decision or service involved.

This is where a connected approach to data strategy, data governance, data quality and AI readiness becomes important. Together, these foundations help agencies move from broad AI ambition to use cases that can be implemented, monitored and evaluated responsibly.

What happens when the underlying data is not ready?

An AI system depends on the information available to it.

If that information is incomplete, duplicated, inconsistent, inaccessible or poorly governed, greater AI capability can amplify existing weaknesses rather than resolve them.

This is why AI readiness is not simply a technology assessment. Before choosing or expanding a tool, an organisation needs to understand:

  • what information it holds and where it sits
  • whether important definitions are consistent across teams and systems
  • whether the data is sufficiently accurate and current for the proposed use
  • who is responsible for its quality, access and governance
  • how information can be used lawfully, safely and transparently
  • whether the intended outcome can be measured using reliable evidence

“Organisations do not need to clean every dataset before they can begin using AI,” Pierre says.

“They do need to identify the business or service problem first and then determine which data supports it.

“That creates a practical path: improve the information that matters, test the technology against a real outcome and learn before scaling.”

Start with the problem, not the AI tool

The OECD’s findings point to five practical questions for leaders responsible for AI, data and digital investment.

1. What outcome are we trying to improve?

Define the operational, organisational or public outcome, not merely the technology to be implemented. An organisation might want to reduce the time people wait for a decision, improve access to a service, help employees find information or identify emerging problems earlier.

Naming that outcome creates a clearer basis for deciding whether AI is the right response.

2. What evidence will show that it worked?

Establish the baseline and measures of success before implementation. Depending on the use case, these measures might include processing time, accuracy, service accessibility, cost, staff workload or user experience.

Without a baseline, organisations may be able to report that a system was introduced but not whether it made a meaningful difference.

3. Is the underlying data fit for this purpose?

Assess the specific information required for the use case, including its quality, completeness, accessibility, timeliness and limitations.

This does not require every dataset across the organisation to be perfect. It requires the organisation to understand whether the information supporting this particular decision or service can be trusted.

4. Who is accountable for the result?

Responsibility must extend beyond technical delivery.Someone needs to own the outcome, the data, the risks and the decision about whether the initiative should be changed, expanded or stopped.

Clear accountability also makes it easier to distinguish between a technical problem, a data problem and a use case that was not well defined in the first place.

5. Can we monitor performance after launch?

AI systems and the conditions around them can change. Measurement therefore needs to continue after deployment, with a way to identify declining performance, unintended consequences and emerging risks.

A system that performed well during a controlled pilot may behave differently when it encounters new information, changing user behaviour or a much larger volume of work.

The next phase of government AI should be judged on results

The OECD outlook does not suggest that governments should stop or delay every AI initiative. It shows why adoption needs to be more disciplined.

Strategies, pilots and infrastructure are necessary, but they are intermediate steps.

The meaningful result is a service that works better, a decision supported by stronger evidence or a process that produces greater value without weakening accountability.

For Australian organisations, the opportunity is to connect AI ambition with the foundations needed to deliver it: a clearly defined problem, trusted and governed data, capable people and measurable outcomes.

The organisations that do this will be in a stronger position to move beyond experimentation and demonstrate what their AI investment has actually achieved.

Is your data ready to support AI?

Before investing in or expanding an AI use case, organisations need to understand whether the supporting data, governance and measurement foundations are in place.

Notitia helps Australian organisations assess their current data environment, define practical use cases and build the strategies, governance and analytics capabilities needed to turn technology investment into measurable outcomes.

Talk to Notitia about assessing your organisation’s data and AI readiness.

References

Frequently asked questions

What is the OECD?

The Organisation for Economic Co-operation and Development (OECD) is an international organisation whose member countries work together on economic, social and public-policy issues. It produces comparative data, research, standards and policy recommendations.

Which countries are members of the OECD?

The OECD has 38 members: Australia, Austria, Belgium, Canada, Chile, Colombia, Costa Rica, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Israel, Italy, Japan, South Korea, Latvia, Lithuania, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, Switzerland, Türkiye, the United Kingdom and the United States.

What is the OECD Digital Government Index?

The OECD Digital Government Index measures how effectively participating governments use digital technology and data to design and deliver public services.

It assesses six areas: digital by design, a data-driven public sector, government as a platform, open by default, user-driven services and proactiveness.

The Digital Government Outlook 2026 found that the average score across participating OECD countries had increased by nine percentage points since 2023. This indicates overall progress, particularly in data governance, service design and AI adoption. However, the OECD found that implementation and evaluation remain significant challenges.

Which countries were covered by the Digital Government Outlook 2026?

The report presents results covering 36 OECD countries and eight countries seeking OECD membership: Argentina, Brazil, Bulgaria, Croatia, Indonesia, Peru, Romania and Thailand.

Not every country necessarily contributed to every indicator, so individual percentages should be read in the context provided by the report.

Does the 75% finding refer to governments around the world?

No. It refers to the OECD countries covered by this research—not 75% of every government in the world.

The OECD found that 75% of the OECD countries studied do not evaluate whether investments in digital technologies and AI deliver their intended results.

Was Australia included in the OECD research?

Yes. Australia was included in the research and has a dedicated country note.

However, the overall 75% result should not be used to claim that Australia does not evaluate its AI investments unless the Australia-specific evidence supports that conclusion.

What does it mean to evaluate an AI investment?

Evaluation means deciding what success should look like, recording the starting position and measuring what changes after the technology is introduced.

Measures might include service-delivery time, accuracy, accessibility, cost, staff workload, user experience and unintended consequences.

Why does data quality matter for government AI?

AI systems rely on the information available to them. Incomplete, inconsistent, outdated or poorly governed data can weaken results and create risks.

Organisations need to assess whether the specific data supporting an AI use case is accurate, accessible, lawful and suitable for the decision or service involved.

What are data-interoperability systems?

Data-interoperability systems allow different departments, organisations and technology platforms to exchange information and use it consistently. They rely on shared standards, definitions and secure methods for connecting systems.

For example, an interoperability system could allow information supplied to one government agency to be securely recognised by another, where authorised, rather than requiring a person to submit the same information again.

The technology may connect the data, but governance is still needed to determine what can be shared, with whom and for what purpose.

Notitia's Data Quality Cake recipe