The NSW Food Authority has issued a tender for an artificial intelligence tool to automate the assessment of penalty notices, aiming to replace a manual process currently requiring a full-time employee earning $115,000 annually.

It is looking for an AI-powered solution to verify the accuracy of penalty notices before publication on the Food Authority's Name and Shame website, with submissions closing 9 February 2026.

The initiative addresses significant compliance and operational challenges within NSW's food safety enforcement system. In the 2024/25 financial year, the Food Authority assessed 1,178 penalty notices issued under the NSW Food Act 2003, with an estimated 40 per cent containing errors or incomplete information requiring manual verification and follow-up with issuing agencies.

The current manual process involves opening each penalty notice record in the Byte regulatory business system, checking scanned documents for accuracy, verifying legal entity details match offence codes and fine amounts from the Fixed Penalty Handbook, and conducting extensive back-and-forth correspondence with 128 NSW councils and enforcement agencies to resolve discrepancies or obtain missing information.

Penalty notices arrive in both handwritten and electronic formats, requiring assessors to verify correct legal entity identification, appropriate offence codes and fine amounts, breach clause citations, complete address information, and premises type classifications before determining publication eligibility through a decision matrix based on priority classification and breach severity.

The Food Authority's Name and Shame website recorded 770,000 hits in 2024/25, creating heightened brand and reputation risks if incorrect information is published. Section 133A-H of the Food Act 2003 mandates the Food Authority publish penalty notice information in good faith and remove improperly served notices from the register.

The proposed AI solution will use image recognition technology to categorise penalty notice data before ingestion into the existing Byte workflow engine, complementing or replacing manual assessment steps where possible. The stand-alone tool will integrate with Byte, which manages food licensing, biosecurity authorisations, inspections, enforcement actions and the Name and Shame register.

Revenue NSW currently provides penalty notice files to the Food Authority's SFTP site weekly, triggering automated imports into Byte. The AI tool aims to supplement this workflow by automatically identifying incorrectly issued notices that cannot be published, flagging notices with missing verification information, and validating complete and correct notices ready for publication.

The successful supplier must train internal specialists to provide ongoing technical support in-house, ensuring knowledge transfer and reducing dependence on external vendors for system maintenance.

Project objectives include ensuring watertight and consistent assessment procedures, providing due diligence assurance for published information, significantly reducing staff time and resource burden, minimising human error risk, delivering cost savings, and creating equitable processes for all offenders.

The project timeline spans 20 February to 19 June 2026, including supplier engagement, scoping workshops, system design and development, testing and pilot implementation.