Procurement reform 2026: what it means for procurement teams and where AI fits in

Latvia is undergoing its largest public procurement reform since 2017. Goods and services thresholds are rising from EUR 10,000 to EUR 140,000. Construction from EUR 20,000 to EUR 5.4 million. The number of mandatory exclusion criteria is dropping from 12 to 2. Procedure duration, according to Ministry of Finance estimates, will shorten by 25%.

The numbers look good. But what happens in practice?

Fewer procedures, more complex tenders

The reform logic is simple: everything below the new EU thresholds moves to simplified procedures. As a result, the number of formal procurement procedures in Latvia will decrease. But those that remain will be larger and more complex.

In 2024, Latvia announced 11,421 procurements with a total contract value of EUR 5.45 billion (excluding Rail Baltica), roughly 13% of GDP. After the reform, smaller procurements will fall outside the Public Procurement Law, but the remaining ones will be exactly those where bid evaluation is hardest: multi-part tenders, complex technical specifications, documentation running hundreds of pages.

This is precisely the segment where Latvia already has a problem. The State Audit Office's December 2024 review found:

The reform doesn't change this. In complex procurements, evaluation quality depends on the people doing it. And Latvia doesn't have enough of them.

The talent shortage isn't going anywhere

Latvia has no formal procurement specialist certification system. The European Commission has recommended one, but it doesn't exist yet. The State Administration School (VAS) offers training based on the EU ProcurCompEU competency framework, but it's voluntary.

In practice, this means procurement team members often learn on the job. When a tender is complex, when proposals run 200 pages, when you need to compare five bidders point by point, human capacity simply runs out. A person gets tired by page 50. Loses attention after four hours. Works worse on Friday evening than Monday morning.

The reform doesn't increase the number of procurement teams. It only reshapes the workload: fewer small procedures, more large ones.

Transparency increases, even below thresholds

The reform's most significant structural change: even below-threshold procurements must now publish procurement plans, concluded contracts, and actual expenditures. The Procurement Monitoring Bureau (IUB) takes over the entire Electronic Procurement System (EIS) and becomes the central authority.

The Ministry of Finance positions this as "data-driven governance", referencing data analytics solutions, reference prices, and KPI-based ex-post monitoring on the reform page.

IUB Director Artis Lapins has publicly acknowledged plans for AI-based procurement data analysis, while noting that "it requires significant funding" and that AI can "generate hallucinations." The skepticism is understandable. But it also reveals that the institutional pace of AI adoption will be slow.

AI in procurement: what Latvia does and what nobody does

In Latvia, the intersection of AI and procurement already exists. iepirkumi.lv offers a training course "Effective use of artificial intelligence in procurement specification and bid evaluation", where specialists learn to use ChatGPT and Claude as personal assistants.

The first can be taught in a three-hour course. The second requires an agent architecture capable of processing hundreds of pages and reasoning about them the way an experienced procurement specialist would.

The State Audit Office's May 2025 review on AI in public administration shows the situation:

Moreover, not a single institution has a mechanism for measuring AI usage costs or results.

What the reform means for tool selection

In the context of the reform, procurement teams need to think about two things.

First, complex tenders are getting more complex. Centralization means municipalities must now manage procurements centrally. New transparency requirements mean more documentation, even if the procedure itself is simpler. Procurement commissions will have more work, not less.

Second, AI tools are not the future — they are the present. The question is not whether to use AI in procurement. The question is how to use it so that every finding is backed by evidence from the document, not by an AI "hallucination."

At Mitigate, we've built an AI agent that reads every page of the tender documents and proposals, compares requirements point by point, and delivers findings with direct quotes from the documents. No guessing. No hallucinations. Every finding is verifiable.

The reform is an opportunity, not an obstacle

Latvia's procurement system processed EUR 5.45 billion in 2024.

These savings are possible only if procurement commissions can identify non-compliance, compare proposals thoroughly, and make decisions based on all available information — not just the first page and the lowest price.

AI agents don't replace procurement specialists. They do what no human can: read everything with equal attention from the first page to the last, every time. And then the expert can focus on what AI cannot: judgment, strategy, and decisions.