Ways to buy AI
| Who | Procurement, project, and technology teams |
| What | Description of AI use cases, aligned with type of AI feature, customization level, primary purchase type, commercial model (pricing and contract type), commercial risks, AI risks, and common procurement methods. |
Governments often purchase AI through three pathways: “Off-the-shelf” for commercially available AI solutions for quick, common needs, “Customize” for proven AI systems to fit specific processes, or “Build” for new AI solutions for strategic, high-stakes, or emerging challenges. Your procurement pathway should be determined by your use case, and informed by other factors such as risks.
Right now, off-the-shelf is the most common way we hear the public sector is buying AI, and this is increasingly happening through central purchasing. This trend implies a more limited role for procurement, both in terms of the purchasing process itself and supporting adoption. Custom or build solutions mean the public sector is purchasing software development services and time, rather than simply a product. This makes good technology procurement practices critical for success.
Not all AI enters the public sector through these pathways. Sometimes AI isn’t deliberately purchased, but is added on to existing services under contract that the public sector is already using. Other times, AI is provided to public sector organizations through academic partnerships, grants, or pro bono pilots with private sector companies.
Do you feel stuck? We’ve heard over and over again that teams don’t feel they have a choice when it comes to which pathway they pursue. If this is you, check out the Hacks for better AI procurement section for ideas on how to get the most out of your procurement if you can't choose the method.
Use cases of procurement pathways
Download the table: Use cases for procurement pathways
Buying off-the-shelf is often pursued when… Use case: Solutions are well-established and standardized ROI: Clear ROI can be more easily demonstrated Capacity: Internal technical resources are limited Data security: No custom solution required to protect data beyond the organization’s standard privacy agreements. Timeline: Quick implementation is needed Vendors: Established vendors exist, are pre-approved already and can deliver quickly
Buying customized or build AI is often pursued when… Use case: Unique organizational needs exist ROI: May be demonstrated or less certain for newer use cases Capacity: Internal technical capability is strong Data security: Custom solutions required Timeline: Sufficient time to accommodate agile development Vendors: Pool of vendors with known capabilities, newer vendors and collaborations are needed