What AI Procurement Really Looks Like in 2025:

In July 2025, over 40 experts from across public, private, and social sectors gathered in Cardiff for the AI Procurement Unconference, hosted by Posterity Global. The event offered an unfiltered, practical look at the real-world challenges public bodies face when buying and deploying AI. What emerged was clear: this is not just a tech problem, it’s a cultural, governance, and capacity problem too.

Key Themes from the Day

1. Language Gaps and Legal Grey Areas

From IP ownership to exit clauses and cybersecurity standards, procurement teams are wrestling with outdated frameworks that don’t reflect the realities of AI. The report calls for clearer definitions and three distinct lenses: buying AI, supplying AI, and using AI in procurement.

2. Massive Knowledge and Skills Deficit

Many public servants and SMEs feel overwhelmed by AI. There's a widespread lack of confidence in understanding how AI works, how to assess tenders, and how to manage AI contracts ethically. Without better pre-market engagement, training, and internal support, public bodies are flying blind.

3. Procurement Is Still Undervalued

Procurement teams are overburdened with manual tasks and low-value purchases, leaving little space for strategic work like AI. The divide between “procuring AI” and “using AI to improve procurement” remains poorly understood.

4. Legal Complexity and Cultural Sensitivity

Public organisations face a moving target of legislation, while also needing to protect cultural identity. In Wales, for example, there are concerns that generic AI tools may erode local culture and fail to support the Welsh language.

5. Legacy Infrastructure as a Major Barrier

The hype around AI often ignores the painful reality of legacy systems. Public bodies simply don’t have the data maturity or infrastructure needed for AI to succeed.

6. Cultural Resistance and Risk Aversion

Fear of failure, siloed thinking, and lack of experimentation are stalling progress. Teams are divided between early adopters and sceptics, and without clear guidance, many choose inaction.

7. Poor Data Quality and Security Risks

Unstructured, outdated data—and fears around IP loss and AI misuse—are major roadblocks. Some departments still allow open AI tools that could inadvertently leak sensitive information.

8. Ethics and Accountability Loopholes

Without human oversight, AI risks biased outcomes and undermines fairness in tenders. Questions around public value, IP rights, and how to audit AI-generated content remain unresolved.

9. Broken Procurement Models

Traditional contracts are ill-suited for dynamic AI tools. There's a high risk of vendor lock-in, unfair competition, and outdated social value measures that don’t reflect AI's impact.

10. AI's Potential for Good

Despite the challenges, there was a strong belief in AI's power to help. Whether enabling better service delivery, supporting SMEs with bids, or freeing up buyer time—AI, when used well, could be a force multiplier for government.

What's Next?

Posterity Global is leading several follow-up actions:

• A Community of Interest, kicking off with an online event on 17 September 2025. See this LinkedIn post for additional information and this link to register.

• In partnership with Apolitical, a Community of Practice for AI procurement professionals in government, to be hosted in their Innovative Procurement Community.

• Continued publication of AI procurement insights and playbooks

Final Thought

This moment requires more than incremental change. Public bodies need a bold, joined-up strategy grounded in data integrity, ethical AI, and adaptive procurement. Done right, AI can deliver real public value—but only if the public sector is ready to meet it on its own terms.