Public procurement is where trust in government is actually tested

If you want to know whether a government can be trusted, don’t read its speeches. Read its contracts.

Public procurement — the buying of everything from school meals to highways — is one of the largest flows of money on Earth. The Open Contracting Partnership estimates governments spend around US$13 trillion a year through contracts; in OECD countries procurement averages about 12–13% of GDP (OCP; OECD). It is also, by wide agreement among anticorruption bodies, the single government activity most vulnerable to corruption and waste — with losses commonly estimated at 10–25% of contract value in affected systems (UNODC; OECD).

We look at procurement not because it is scandalous but because it is structural: it is millions of consequential decisions, made by identifiable people, on evidence that can be published. It is, in other words, exactly the shape of problem trustworthy decision intelligence exists for.

Every contract is a chain of decisions

Strip away the paperwork and a procurement is a sequence of judgments: Is this purchase necessary? Is the specification written for the need, or for a favored vendor? Is this bid genuinely the best value? Is the price anomalous against the market? Was delivery actually what the contract promised?

Each judgment is a place where good analysis changes outcomes — and where the absence of visible reasoning is what lets bad outcomes hide. A specification tailored to one bidder looks, from the outside, exactly like a specification: the difference only appears when you can compare it against hundreds of similar purchases. That comparison is an evidence problem, and until recently it was too expensive to do routinely.

What happens when the decisions become inspectable

The natural experiment has been run. Ukraine’s ProZorro system, built after 2014, published the full procurement process as open data by default; independent assessments and the government’s own figures credited it with saving on the order of a billion dollars within its first years, alongside a measurable rise in bidder participation (open-contracting.org). Colombia’s SECOP platform made national and territorial contracting searchable by any citizen, journalist, or oversight body (Colombia Compra Eficiente); Paraguay, Chile, and a growing list of countries publish to the same open standard. The pattern across cases is consistent: participation rises, single-bid awards fall, and watchdogs — official and civic — catch what they could never previously see.

Transparency alone did that. Which raises the question we find most interesting: transparency makes the decisions visible after the fact. What improves the decisions before they are made?

The next layer: evidence at the moment of decision

A procurement officer in a mid-sized city — Barranquilla, Kraków, Cebu — deciding whether a bid is sound has, typically, the documents in front of them and their own experience. The open data that could inform them (comparable purchases, market prices, each supplier’s delivery history, red-flag patterns) exists, but assembling it for one decision takes analytical staff that most agencies do not have.

This is precisely where AI-assisted analysis is now being aimed — and precisely where the confident-answer problem becomes dangerous. An opaque model flagging a bid as “risky,” with no evidence attached, is an accusation no official can responsibly act on and no supplier can fairly contest. Held to the Standard, the same capability becomes usable: this price is 34% above the median of 212 comparable purchases — here they are; this supplier’s delivery record is here; this is what the flag assumes; here is the confidence; a named analyst reviewed it; here is how to challenge it.

The difference is not the algorithm. It is whether the answer arrives as a verdict or as evidence.

Why this problem is worth years

Procurement sits at the intersection of everything this institution cares about: public money at planetary scale, decisions concrete enough to support with real evidence, data that is increasingly open, and a trust deficit that transparency has narrowed but not closed. Progress here is measurable in the most unsentimental terms available — prices paid, competition levels, delivery rates — which makes it a domain where a decision-intelligence institution can be checked against reality rather than taken on faith.

That is our kind of problem. The money is public. The decisions should be legible to the public they serve.

Sources

• Open Contracting Partnership — global procurement spend estimates, ProZorro and country case studies. open-contracting.org
• OECD — Public procurement (size, governance, and integrity data for member countries). oecd.org
• UNODC — guidance and estimates on corruption in public procurement. unodc.org
• Colombia Compra Eficiente — SECOP open contracting platform. colombiacompra.gov.co