As public agencies globally turn to artificial intelligence to modernize services, procurement processes are facing unprecedented scrutiny. The Open Contracting Partnership (OCP) argues that many governments are moving too fast, adopting tools they do not fully understand and cannot effectively deploy. Its new guide, Buying AI: Tools and Tips for Public Procurement, urges stronger due diligence and earlier cross-functional involvement to ensure AI investments are grounded in real capability and need.
The pressure to show progress, combined with limited internal technical skills, has led to poorly scoped projects, over-reliance on vendor claims, and limited evaluation of long-term outcomes. According to OCP, these dynamics are particularly risky in procurement and supply chain functions, where tools promising automation and risk intelligence can underperform when deployed without proper operational planning.
The Cost of Rushed AI Adoption
AI pilots in government are expanding rapidly, from case-processing automation to predictive maintenance and digital citizen services. But speed has come at a price. Public agencies without dedicated AI oversight structures are frequently procuring solutions that fail to scale or integrate with legacy systems. OCP, active in over 50 countries, notes that teams often lack the expertise to interrogate performance claims or assess ethics, data security, and accountability risks.
“Governments are racing to adopt AI, but too often they’re doing it without the preparation and guardrails to make those investments succeed,” says Kaye Sklar, Senior Programme Manager for Content and Insights at OCP.
Recent high-profile examples reinforce the caution. The UK’s National Audit Office has called for clearer value-for-money frameworks as generative AI investments increase across departments. And in the U.S., the White House AI executive order highlights testing, bias safeguards, and procurement guidance for federal agencies, reflecting concern that untested tools could undermine services or public confidence.
Procurement functions are central to avoiding these pitfalls. OCP stresses early involvement so purchasing teams can shape requirements and ensure contracts demand measurable outcomes, rather than reacting to pre-selected technologies after decisions are made. Without procurement at the table from the outset, agencies risk buying the wrong tool for the wrong problem.
Proof Points From Early Adopters
A growing set of government programs demonstrates what disciplined adoption looks like. At the U.S. Centers for Medicare & Medicaid Services, AI-enabled service platforms developed with Skyward IT Solutions are supporting internal operations and procurement analytics, contributing to monthly savings of roughly US$3 million through faster ticket processing and spend insights. In Australia, the Digital Transformation Agency maintains a national playbook of standards, policies, and model contracts to streamline AI adoption while enforcing consistency and transparency. Chile’s digital agency has gone a step further by creating an AI “sandbox” to test vendor tools before deployment, improving validation and reducing procurement risk.
This structured rigor reflects wider momentum. The OECD’s AI Principles and the EU’s AI Act both emphasize procurement controls, transparency, and accountability. According to trade reports, cities deploying AI for mobility, waste tracking, and infrastructure monitoring increasingly require vendors to pass technical audits and share real-world performance metrics.
“Poor AI procurement can cost governments millions, not just in failed projects, but in lost public trust,” says Kathrin Frauscher, Deputy Director at OCP.
To institutionalize good practice, OCP outlines core roles, procurement leads, project owners, data experts, and legal counsel, who must collaborate from initial scoping to post-implementation evaluation. In supply chain contexts, this approach enables more reliable forecasting, stronger vendor oversight, and clearer standards for algorithmic decision-making.
The Next Procurement Competency
As AI matures, the advantage won’t come from speed of adoption but from institutional learning loops that turn policy into muscle memory. Some governments are already building this discipline: Canada’s Treasury Board, for example, requires algorithmic impact assessments before deployment, and Singapore’s Model AI Governance Framework ties procurement criteria to risk classification and auditability. These approaches suggest a shift underway, procurement is becoming a core capability in algorithmic governance, where success depends not just on selecting the right tools, but on developing the internal judgment to scrutinize them continuously.
