19 Aug 2026

IM BLOG: Beyond the Prompt: Preparing for Agentic AI in the Australian Public Sector

Over the past year, generative AI tools such as ChatGPT and Claude have begun to influence how public sector organisations access and use information.

Over the past year, generative AI tools such as ChatGPT and Claude have begun to influence how public sector organisations access and use information. While these tools offer productivity gains, they remain fundamentally reactive, responding to prompts but not acting independently.

A more significant shift is now emerging.

Agentic AI refers to systems that can not only generate responses but also take action, interacting with systems, applying rules, and executing workflows. For Records and Information Management (R&IM) professionals in the Australian public sector, this development has direct implications for:

Freedom of Information (FOI) processes
records governance and accountability
compliance with legislative frameworks
public trust and transparency

From Responding to Acting

Traditional AI tools sit within a defined interaction:

a user submits a query
the system generates a response

Agentic AI extends beyond this by:

locating and retrieving records across systems
applying classification and retention rules
initiating workflows (e.g. disposal, access, escalation)
documenting decisions and actions taken

This represents a shift from ‘human-in-the-loop’ to ‘human-on-the-loop’, where professionals oversee systems rather than perform each task manually.

In a public sector context, this transition raises a critical consideration:

If an automated system acts on information, who is accountable and how is that action demonstrated?

Implications for FOI and Access to Information

FOI functions are particularly sensitive to this shift.

Agentic AI could be used to:

search across multiple repositories to locate responsive documents
collate and prepare FOI responses
identify exempt or sensitive information
support decision-making processes

While this introduces efficiency, it also raises risks:

Completeness risk: Was every relevant record identified?
Bias or exclusion risk: Were sources prioritised appropriately?
Defensibility risk: Can the agency demonstrate how a decision was reached?

Under Australian FOI obligations, agencies must demonstrate that reasonable steps were taken to locate documents and that decisions are transparent and reviewable.

If AI is involved in these processes, its actions must be:

traceable
explainable
auditable

1. Coordinated AI Systems and Recordkeeping Integrity

A key development in Agentic AI is the use of multiple specialised systems working together, rather than a single general-purpose tool.

In practice, this could involve:

one system identifying relevant records
another applying retention or disposal rules
another verifying compliance with policy

This aligns with structured recordkeeping practices, but introduces a new requirement:

Each step in the process must remain visible and defensible.

For R&IM professionals, this reinforces the importance of:

clearly defined classification frameworks
consistent metadata standards
controlled workflows with audit trails

2. AI Acting Within Government Systems

Agentic AI is increasingly being integrated with enterprise platforms, document and records management systems (EDRMS), case management systems, and data repositories.

This creates new capabilities, such as:

automated application of retention schedules
proactive identification of high-risk or sensitive information
faster response to audits, inquiries, and FOI requests

However, it also amplifies risk:

actions can occur rapidly and at scale
errors may propagate across systems
unintended disclosures may occur if controls are insufficient

This makes governance-by-design essential.

R&IM professionals must ensure that:

policies are embedded into system behaviour
access controls and permissions are maintained
automated actions align with regulatory obligations

3. Auditability and Legal Defensibility

In the Australian public sector, decisions relating to records must be capable of withstanding scrutiny, from auditors, regulators, tribunals, and courts.

As AI becomes more involved in decision-making, organisations must be able to demonstrate:

how a decision was made
what information was used
why a particular outcome was reached

This is especially important in:

FOI reviews (e.g. by the Information Commissioner)
administrative law challenges
data breach investigations

Emerging AI practices, including decision tracing and detailed activity logging, support this requirement.

For R&IM, this reinforces a fundamental principle:

If a decision cannot be explained, it cannot be defended.

4. Evolving Information Retrieval in Investigations and FOI

AI-driven retrieval is also evolving beyond simple search.

More advanced approaches can:

synthesise information across multiple systems
refine results iteratively
identify gaps or inconsistencies in records

This capability has clear benefits for:

FOI processing
internal investigations
audit preparation

However, it requires careful governance to ensure:

only authorised data sources are accessed
sensitive information is handled appropriately
outcomes are reproducible and verifiable

The Expanding Role of R&IM in the Public Sector

The emergence of Agentic AI extends the role of R&IM professionals beyond traditional recordkeeping functions.

The focus is shifting from:

Managing records

Governing how automated systems use and act on information

This creates a leadership opportunity for R&IM to contribute to:

AI governance frameworks
information risk management
compliance assurance in automated environments

Importantly, R&IM provides the structure and controls needed to ensure that AI operates in a way that is:

accountable
transparent
compliant with legislative obligations

Conclusion

The evolution from chatbot-based tools to agentic systems marks a significant transition for the Australian public sector.

While conversational AI will continue to play a role, the increasing ability of systems to act on information requires a stronger emphasis on governance, oversight, and accountability.

For R&IM professionals, this is not a peripheral issue, it sits at the core of public sector responsibilities.

As agencies explore these technologies, the challenge is not only to improve efficiency, but to ensure that:

decisions remain defensible
records remain trustworthy
and public confidence is maintained

In this environment, R&IM is not just supporting the organisation, it is helping to safeguard its integrity.

Meet your blog author:

Return to listing