The surprising shifts in how the public sector is buying AI, and what policymakers can do about it
Over the past year, we at Open Contracting Partnership spoke to over fifty public sector practitioners and experts across the United States, Europe, and beyond to understand how the public sector is buying AI. Here is what we learned from the frontlines.
Buying AI today: what’s really happening
Public sector organizations are accelerating their investments in AI technology, and spending big: In the UK, government contracts for AI projects hit £573 million by August 2025, exceeding all of 2024. In the United States, federal agencies committed $5.6 billion to AI between 2022 and 2024. But it’s not just what they buy, it’s how they buy it that will have a huge impact on outcomes.
1. Off-the-shelf AI is winning over custom builds.
Organizations aren’t rushing to buy complex, custom-built AI systems. Instead, right now they are purchasing off-the-shelf licenses for lower-risk efficiency-driven use cases, such as AI-powered writing assistants, data analysis tools, or automated document management systems. Public sector organizations can often use these tools through their existing cloud or productivity platforms.
2. Centralized buying is on the rise.
We see a clear shift toward enterprise-wide AI procurement. Central IT or digital transformation agencies now negotiate contracts for all government departments. The United States, among others, has moved to this model. While central purchasing can promote efficiency and interoperability, this also means that decision-making power is concentrated in fewer hands.
3. AI is sneaking in through side doors.
Not all AI used by the public sector goes through procurement. Government agencies often access AI through free pilots, grants, features built into existing tools, or academic partnerships. This “shadow AI” can help teams move fast, but it means less opportunity for accountability and oversight.
Together, these trends create a growing gap between AI procurement and AI adoption. IT teams manage enterprise licenses for AI, but how these AI tools are used is left up to individual agencies. Public servants are also implementing AI tools that their agencies never formally purchased. These approaches can be efficient for getting AI tools into the hands of public servants, but can pose challenges during implementation, such as a lack of accountability in terms of use, increased dependency on vendors, and a general lack of capacity to apply these tools properly.
What this means for policymakers
To ensure that the rollout of new AI tools and services goes well, policymakers must focus on increasing their organization’s AI readiness. Here’s what this looks like in practice:
1. Start with vision and governance.
Recent polling of public servants found that over 60% have experimented with AI, but only 35% had received any guidance. Our research found that AI-ready public sector organizations have an organization-wide strategy that clearly outlines how they envision using AI technology to achieve specific priorities, and also identifies frameworks for governance, transparency, and accountability, and their plan for implementation. This approach allows organizations to take advantage of the new opportunities created by AI, while managing risk. For example, the City of Seattle’s new 2025-2026 AI Plan identifies specific, high-ROI AI use cases that the city will focus on for the coming two years, a roadmap for implementation, a plan for upskilling staff, governance mechanisms, and the infrastructure requirements to carry it all out.
2. Build in-house AI capacity.
Buying AI is not enough; governments must understand it. Organizations must build their in-house expertise to evaluate tools, design responsible systems, manage vendors, and use the technology that they buy. Right now, far more needs to be done. A recent Government AI Landscape Assessment by Code for America found that US states performed the worst in AI capacity building, with most still in the early stages. In the United States, organizations like GovAI Coalition and InnovateUS are working to fill this gap.
3. Make procurement a driver of innovation.
As governments shift to central procurement of technology, success depends on how well these tools are integrated and adapted to real user needs. Policymakers should empower procurement teams to buy not just products, but agile, design-driven service teams that can build ethical, flexible, and future-ready systems. This is especially true when buying more complex customized solutions, where AI can potentially deliver high value for the public sector. Procurement teams must be allowed and encouraged to buy in more agile and collaborative ways that support testing and iteration, and that open the door to working with smaller vendors that don’t typically participate in public procurement. For example, the State of Georgia issued a Request for Qualified Contractors (RFQC) to identify a pool of service providers to partner with agencies to solve their problems by leveraging AI – and now the state has a strong pool of nineteen providers, many of whom are smaller vendors who work less with government.
With public sector organizations under pressure to make the most out of every taxpayer dollar, AI presents an opportunity to increase efficiency. But it’s not enough to buy AI solutions—organizations must plan to be ready and integrate them well. Our new guide dives deeper into these insights and offers practical steps, tips and templates, and some good examples for procurement directors and policymakers who want to make this happen. Take a look and see if it can help you!
