AI ambition meets enterprise reality
NAB had already begun exploring how AI could improve internal operations and support employees in navigating complex information environments.
Teams were actively evaluating use cases and experimenting with tools—but quickly encountered a familiar constraint.
The challenge was not the capability of AI systems. It was the condition of the organisation’s knowledge environment.
The challenge: information exists, but cannot be reliably retrieved
NAB operated across a complex landscape of systems, with SharePoint playing a central role in document storage and collaboration, alongside other internal platforms containing specialised operational knowledge.
From a systems perspective, the organisation had the information it needed. Policies, procedures, and guidance were documented and accessible.
In practice, however, retrieving the right information consistently was difficult.
Employees could search across systems, but determining which sources were current, authoritative, and relevant often required experience rather than system support. Information was distributed across repositories with varying structures, ownership models, and update cycles.
This created two related challenges.
First, staff spent time navigating systems and validating information rather than applying it. Second, early AI initiatives struggled to produce reliable outputs because they lacked a consistent way to retrieve trusted knowledge.
The issue was not the absence of information. It was the absence of a reliable retrieval layer.
The approach: establish governed retrieval across systems
Rather than starting with AI interfaces or automation, the focus shifted to improving how knowledge could be accessed across existing systems.
The objective was to create a governed retrieval layer that could:
- Connect knowledge across SharePoint and internal platforms
- Respect permissions and access controls
- Surface information based on relevance and authority
- Provide a consistent foundation for both search and AI
This approach did not require replacing existing systems or restructuring content. Instead, it introduced a layer that allowed those systems to function as part of a unified knowledge environment.
Governance was a central requirement.
Any retrieval capability needed to operate within the organisation’s existing security model, ensuring that users could only access information they were permitted to see. This was essential in a regulated environment where auditability and control are non-negotiable.
What changed: from fragmented access to trusted retrieval
With a governed retrieval layer in place, the organisation gained a more consistent way to access internal knowledge.
Employees were able to locate relevant information without navigating multiple systems or relying as heavily on informal knowledge. More importantly, they could do so with greater confidence in the authority of the information they retrieved.
This reduced friction in day-to-day work and improved the usability of existing knowledge assets.
It also changed how AI could be approached.
Because retrieval was grounded in governed, approved sources, it became possible to explore conversational interfaces and AI-assisted workflows without bypassing existing controls. Responses could be linked back to source documents, allowing users to verify information and maintain trust in the system.
In practical terms, this meant less time spent searching across systems and validating sources, and more time applying information to real work. It also reduced reliance on informal knowledge and workarounds that had previously filled gaps in access.
Key outcomes
- Faster access to trusted information
- Reduced reliance on informal knowledge
- Improved consistency in how information is used
- Foundation for governed AI and conversational access
Why this matters in a regulated environment
In financial services, the introduction of AI is not simply a question of capability. It is a question of control.
Systems must operate within strict governance frameworks, respect permissions, and provide visibility into how information is accessed and used. Without these foundations, AI initiatives risk producing outputs that cannot be trusted or adopted in practice.
By establishing a governed retrieval layer first, NAB created the conditions for AI to be introduced in a way that aligned with these requirements.
AI did not need to be treated as a separate or experimental layer. It could be built on top of an existing, controlled knowledge environment.
From experimentation to operational foundation
The most significant shift was not the introduction of a new interface or tool. It was the improvement in how knowledge was retrieved and trusted across systems.
This created a foundation that supports both immediate operational improvements and future capabilities.
Search became more reliable. Information became easier to access. And AI initiatives could move forward in a way that was grounded in the organisation’s existing governance model.
Establishing governed retrieval did not complete the organisation’s AI journey.
It made it possible to move forward with AI in a way that was controlled, reliable, and aligned with how the organisation already operated.
Closing reflection
Establishing governed retrieval did not complete the organisation’s AI journey.
It made it possible to move forward with AI in a way that was controlled, reliable, and aligned with how the organisation already operated.
