Procurement and GenAI
Like any part of business, procurement of services and systems, planning, following best practice and working to mitigate any risk is critical. GenAI procurement is no different.
This guidance is aligned with the following OECD AI principles:
The first consideration is whether AI is the best solution for your agency. Do this by:
- Planning appropriately — assess business need, analyse the market and factor in risk mitigations.
- Developing robust evaluation criteria — be explicit about pre-conditions and include an exit strategy.
- Doing your research — know the market, suppliers in it and their offerings.
- Using guidance — to help you get the best result, including supporting market resilience.
AI systems present opportunities for government to modernise and streamline operations and improve the delivery of public services. You should make sure you’ve planned appropriately for the size, risk and complexity of your procurement. AI system and service procurement may require a higher degree of risk planning to identify and mitigate the risks it may pose.
Plan for your business needs
AI products differ widely, and AI may not always be the best solution. While something is free at the outset, there may be cost incurred later on once the solution is scaled. Think carefully about your evaluation criteria and if commercial protections might be needed in your contracts. Consider other key foundations, such as privacy and security, when developing evaluation criteria.
Procurement teams should conduct thorough market research on suppliers and their offerings. Consider supplier reputation, capability, pricing, contract value and potential risks in their supply chain.
You’ll get a better and more assured outcome if you consider every phase of the procurement cycle in your planning and manage the contract throughout its life. You’ll need to consider how the Government Procurement Rules apply and keep the Government Procurement Principles front of mind during your planning.
Rules: Government Procurement — NZ Government Procurement
Principles: Government Procurement — NZ Government Procurement
Follow procurement best practice
The Government Procurement Framework is flexible and can accommodate the fast-paced innovative nature of AI. This flexibility depends on following procurement best practice and incorporating adequate risk assessment and mitigation in your procurement plan. There are several domestic and international best practice guides for procuring AI. You should familiarise yourself with these before approaching the market.
Principles, charter and rules — NZ Government Procurement
Early in your planning, conduct a robust needs analysis and consider your specific business need and the role AI is needed to perform. Think about how the GenAI system will meet the needs of your agency, and how it will integrate into existing systems.
The procurement of AI may provide opportunities to use innovative procurement approaches and practices that can help you get the best outcomes from your AI procurement.
Leverage interactive procurement processes that can help you get the best outcomes from your AI procurement, for example:
- Lean Agile Procurement
- Procurement-In-A-Day
- Competitive Dialogue
This helps support a diverse, competitive, and resilient AI supplier market.
Resources for procurement leads
Learn more on the NZ Government Procurement website:
Procurement good practice guides
- AI Procurement Guides — AI Forum New Zealand
- AI Procurement Guide — Tony Blair Institute for Global Change (PDF 109KB)
- Workbook — Unlocking Public Sector AI: AI Procurement in a Box — World Economic Forum (PDF 7.1MB)
- The AI-RFX Procurement Framework — The Institute for Ethical AI & Machine Learning
- Guidelines for AI Procurement — UK Government (PDF 11.1MB)
- Whitepaper — AI in Government Procurement — Australasian Procurement and Construction Council (PDF 14.1MB)
- Buying AI, is the public sector equipped to procure technology in the public interest? — Ada Lovelace Institute (PDF 526KB)
Related guidance
Utility links and page information
Last updated