The OECD.AI Policy Navigator

Our policy navigator is a living repository from more than 80 jurisdictions and organisations. Use the filters to browse initiatives and find what you are looking for.

BI Dave - Natural Language Processing Assistant


-
Added by:   OECD analyst
Added on:   27 Oct 2026
Updated by:   OECD analyst
Updated on:   27 Oct 2026

‘BI Dave’ is an internal AI initiative that helps AusIndustry Insights staff analyse and summarise large volumes of written business information. It supports reporting to senior decision‑makers, including APS portfolios and Ministers, by reducing the time needed to produce clear, consistent briefings. The initiative addresses the problem of slow, manual analysis while keeping people responsible for final decisions.

Initiative overview

‘BI Dave’ is an internal initiative created to help staff make better use of large volumes of written information collected through everyday interactions with businesses. Before ‘BI Dave’, understanding this information required staff to manually read, sort and summarise thousands of notes and reports. This was slow, resource‑intensive, and made it difficult to respond quickly, especially during busy periods or when urgent advice was needed. As a result, valuable insights were often under‑used.

The initiative was developed to improve how efficiently insights could be turned into reports and briefings without relying on more staff. Since going live in March 2024, ‘BI Dave’ has been used frequently and has reduced report drafting time by more than 50 per cent, improving the speed and consistency of reporting. This enables faster, more evidence-based policy and program decisions for government and ensures public feedback is integrated more consistently. Impacts have been measured through before‑and‑after comparisons of drafting effort, operational testing, and staff feedback during routine use. This has also enabled analysis of large volumes of written information that were previously impractical to process manually. Further efficiency and insight gains are expected as similar capabilities are applied to other information sources across the department.

Key challenges included inconsistent data quality, privacy risks from identifying information in free‑text records, and the effort required to prepare data for analysis. Early testing showed variable accuracy across large and mixed datasets, including missed identifiers and occasional mis‑categorisation, requiring controls to prevent inappropriate outputs.

These risks were mitigated through staged, batch‑based testing using a “test–train–repeat” approach, with manual review of outputs and incremental refinement before broader operational use. The initiative highlighted the importance of human oversight, strong quality assurance, and active involvement of business users. Lessons learned include starting with a focused use case, accepting iterative improvement, and embedding governance early. Success depends on reliable internal data, clear accountability, skilled reviewers, and organisational support for AI as an assistive capability rather than an automated decision‑maker.
 

Other relevant details

‘BI Dave’ shows how a small, carefully designed initiative can deliver strong results when built around real business needs. It uses the department’s own information and produces outputs that align with existing reporting styles and requirements. Importantly, it does not replace human judgement—staff review, refine and approve all outputs—helping maintain trust, accountability and quality. Over time, ‘BI Dave’ is expected to continue as a standard part of day‑to‑day operations. Its design allows it to be improved gradually and adapted to support other types of written information, such as data from call centres or grants programs. This creates opportunities to extend the benefits more broadly across the department, using the same approach to improve efficiency, insight and responsiveness.

About the policy initiative