Tēnā koutou,

Pursuant to section 5 of the Official Information Act 1982 (OIA), I request the release of all information held by the Accident Compensation Corporation (ACC) relating to the acquisition, deployment, governance, evaluation, and planned expansion of artificial intelligence (AI) tools and technologies across the entire organization. This request is motivated by the public interest in transparency regarding ACC's adoption of AI, particularly in light of the New Zealand Government's Public Service AI Framework (released 29 January 2025) and the New Zealand AI Strategy (launched 8 July 2025), which emphasize responsible, human-centered AI use to enhance public services while safeguarding privacy, equity, and accountability.

As New Zealand's primary no-fault injury compensation provider, ACC's use of AI in areas such as claims processing, customer service, risk assessment, and data analytics has significant implications for claimants' rights, operational efficiency, and compliance with Te Tiriti o Waitangi principles. Recent disclosures from other agencies, such as the Crown Law Office's OIA response dated 9 October 2025 (available at https://fyi.org.nz/request/32436), highlight the need for comprehensive public oversight of AI integration in government entities.

To ensure clarity and specificity, this request is structured into thematic sections with numbered questions. Where applicable, please provide information in electronic format (e.g., PDFs, spreadsheets, or datasets) for the period from 1 January 2020 to the date of this request (12 October 2025), unless otherwise specified. If any part requires clarification, please contact me promptly. I am happy to receive information in summarized or aggregated form where it reduces volume without losing substantive detail, and I consent to any reasonable charges for extensive copying or collation, provided they are notified in advance.

Please acknowledge receipt of this request within 5 working days and provide a decision no later than 20 working days from receipt, as per sections 15 and 23 of the OIA. If extensions are needed, please justify them under section 15A.

Section 1: Current AI Tools, Subscriptions, and Licenses
1. A list of all AI tools, software, platforms, or services (including generative AI, machine learning models, predictive analytics, natural language processing, computer vision, or automation tools) for which ACC has purchased subscriptions, licenses, or enterprise agreements since 1 January 2020, including:
a. The full name and version of each tool;
b. The vendor or provider (e.g., Microsoft, Google, OpenAI);
c. The date of initial purchase, subscription start, or license activation;
d. The number of licenses or user seats allocated;
e. Any integrations with existing ACC systems (e.g., claims management software, CRM, or data warehouses).

2. For each tool listed in Question 1:
a. The annual license or subscription cost (in NZD, including any GST);
b. Any one-off setup, implementation, or customization fees paid;
c. Breakdown of costs by category (e.g., per-user, enterprise-wide, cloud hosting);
d. Total expenditure to date, including renewals and escalations.

3. Details of any free-tier, trial, or open-source AI tools in use at ACC since 1 January 2020, including:
a. Names and purposes;
b. Dates of adoption;
c. Number of users or departments involved;
d. Any associated indirect costs (e.g., staff time for setup).

4. Copies of all contracts, terms of service, or service level agreements (SLAs) for the top five most expensive or widely used AI tools identified in Question 1, redacting only commercially sensitive pricing if necessary under section 9(2)(i) of the OIA.

Section 2: Deployment and Usage Across Business Units and Departments
5. An organizational breakdown of AI tool usage across all ACC business units, branches, and departments (e.g., Claims, Customer Service, Rehabilitation, Strategy, IT/Digital, Legal, Māori Strategy and Partnerships), including:
a. Which tools are deployed in each unit;
b. Number of active users per tool per unit (as of the latest available date);
c. Primary use cases (e.g., claims triage, fraud detection, customer chatbots, predictive modeling for recovery outcomes);
d. Percentage of workflows or processes automated or augmented by AI in each unit.

6. Quantitative data on AI utilization since 1 January 2023, including:
a. Total number of AI-generated outputs (e.g., decisions, reports, recommendations) per tool per quarter;
b. Metrics on adoption rates (e.g., % of staff trained vs. % actively using AI);
c. Any dashboards, reports, or KPIs tracking AI performance across departments.

7. Examples of AI applications in sensitive areas, such as:
a. Claims assessment or denial processes (e.g., automated eligibility scoring);
b. Customer interactions (e.g., AI chatbots or virtual assistants);
c. Risk profiling or fraud detection (e.g., anomaly detection models);
d. Personalized rehabilitation plans (e.g., predictive analytics for recovery timelines). Provide anonymized case studies or process flows if full details are withheld.

8. Information on AI use in Māori-specific or culturally responsive services, including:
a. Tools adapted for te ao Māori principles (e.g., data sovereignty compliance);
b. Involvement of Te Puni Kōkiri or iwi partners in AI design or evaluation;
c. Any disparities in AI outcomes for Māori claimants (e.g., bias audits).

Section 3: Trials, Pilots, and Evaluations
9. A register of all AI trials, pilots, or proof-of-concept projects conducted by ACC since 1 January 2020, including:
a. Project name, objectives, and scope;
b. Start and end dates;
c. Participating departments and external partners;
d. Tools or technologies tested;
e. Total costs incurred (broken down by category).

10. For each trial/pilot in Question 9:
a. Key performance indicators (KPIs) measured (e.g., accuracy, efficiency gains, error rates);
b. Results and outcomes (e.g., success metrics, lessons learned);
c. Any evaluation reports, including internal reviews or third-party audits;
d. Decisions on scaling up, discontinuation, or modification.

11. Details of any failed or discontinued AI initiatives, including:
a. Reasons for failure (e.g., technical issues, ethical concerns, cost overruns);
b. Costs sunk and any remediation measures;
c. Changes to policies resulting from these experiences.

12. Copies of all internal memos, emails, or meeting minutes discussing AI trial outcomes from the past 12 months, redacting personal information under section 9(2)(a) if necessary.

Section 4: Costs and Budgeting
13. A comprehensive breakdown of ACC's total expenditure on AI tools, infrastructure, and related activities since 1 January 2020, categorized by:
a. Software licenses and subscriptions;
b. Hardware/cloud computing (e.g., GPU resources for model training);
c. Consulting, development, or integration services;
d. Training and upskilling programs;
e. Ongoing maintenance and support.

14. Budget allocations for AI in the current and next financial year (2025/26 and 2026/27), including:
a. Line-item details from ACC's annual plan or ICT budget;
b. Projected ROI or cost savings from AI deployments;
c. Any funding sourced from external grants (e.g., MBIE AI innovation funds).

15. Cost-benefit analyses conducted for major AI implementations (e.g., those exceeding $50,000), including assumptions, methodologies, and outcomes.

Section 5: Policies, Governance, and Risk Management
16. Copies of ACC's internal AI policies, guidelines, or frameworks (e.g., AI ethics charter, governance deck), including:
a. Alignment with the Public Service AI Framework and OECD AI Principles;
b. Processes for AI procurement approval and vendor selection;
c. Requirements for human oversight in AI decisions (e.g., "human-in-the-loop" protocols).

17. Details of AI governance structures at ACC, such as:
a. Committees, working groups, or roles responsible for AI oversight (e.g., AI Ethics Board);
b. Reporting lines to executive leadership or the Board;
c. Frequency of AI risk reviews or audits.

18. Risk management practices for AI, including:
a. Assessments for bias, fairness, and discrimination (e.g., using Stats NZ's Algorithm Impact Assessment tool);
b. Privacy impact assessments under the Privacy Act 2020 for AI data processing;
c. Incident logs of AI-related issues (e.g., hallucinations, erroneous decisions) since 1 January 2023, anonymized.

19. Compliance reporting on AI use, such as:
a. Submissions to the Government Chief Digital Office (GCDO) or MBIE on AI maturity;
b. Adherence to the Algorithm Charter for Aotearoa;
c. Any findings from internal or external audits on AI risks.

Section 6: Training, Capability, and Workforce Impact
20. Programs for building AI capabilities at ACC, including:
a. Training courses, workshops, or certifications provided to staff (e.g., prompt engineering for GenAI);
b. Number of staff trained per department since 1 January 2023;
c. Budget allocated to AI skills development;
d. Partnerships with external providers (e.g., universities, TechHub).

21. Impact of AI on ACC's workforce, such as:
a. Job roles affected (e.g., automation of administrative tasks);
b. Upskilling or redeployment initiatives;
c. Any collective agreements or consultations with unions on AI-induced changes.

Section 7: Future Uses, Plans, and Strategic Integration
22. Strategic plans for AI expansion at ACC through 2030, including:
a. Roadmap for new tool acquisitions or upgrades;
b. Intended use cases (e.g., AI in predictive healthcare analytics);
c. Projected budget and staffing needs.

23. Upcoming procurements or RFPs for AI technologies, including:
a. Timelines and scopes;
b. Evaluation criteria (e.g., ethical AI standards);
c. Shortlisted vendors.

24. Integration plans with national initiatives, such as:
a. Collaboration with other agencies (e.g., Health NZ on shared AI for injury data);
b. Participation in GCDO's AI sandboxes or innovation challenges.

Section 8: Data, Analytics, and Broader Impacts
25. Datasets or models developed or used by ACC for AI purposes, including:
a. Descriptions of key datasets (e.g., claims history for training models);
b. Data governance protocols (e.g., anonymization, Māori data sovereignty);
c. Any open-sourcing or sharing of AI models.

26. Analytics on AI's impact on ACC outcomes, such as:
a. Improvements in claims processing times or accuracy;
b. Claimant satisfaction metrics related to AI interactions;
c. Equity analyses (e.g., outcomes by demographics).

27. Any research, white papers, or studies commissioned by ACC on AI's role in injury compensation, including executive summaries.

Kind regards,

Spencer Jones