Managing the 'weird colleague': AI, procurement and the new skills governments need

Every year, governments around the world spend $13 trillion dollars on procurement — buying the goods, services and technologies they need to function. It’s an astonishing figure — equivalent to around 10% of global GDP. Even more so when you consider that, for too long, procurement has been one the most overlooked parts of government.

“No child ever dreams of becoming a procurement officer,” says Leonardo Quattrucci — Adjunct Professor at Sciences Po, former manager in emerging technologies at AWS and, before that, public servant working for the President of the European Commission.

But the world has changed. In a more globally competitive environment, where more governments are fretting about technological change, innovation and independence, policymakers are taking a closer look at that $13 trillion and asking: What more can they get from it?

More governments are professionalising procurement as a function — although they are still in the minority. The research field is growing and evolving. Researchers are reimagining procurement as a tool for innovation, rather than a purchasing function. Through the Covid-19 pandemic, citizens have a more tangible sense of the very real world implications when governments get procurement right, or wrong.

Leonardo Quattrucci has been thinking deeply about the future of innovation policy and arguing that procurement plays a central role in it. Most notably, he authored a white paper on ‘A Well-Architected Framework for public procurement’, drawing lessons from technology to re-architect public institutions. Apolitical spoke to Quattrucci in May 2025.

A user-centred approach to procurement

Apolitical: Can you describe some of your experiences of managing public procurement processes – what key lessons did you learn?

Leo Quattrucci: During the outbreak of the pandemic, in early 2020, I was the global account manager for an international agency which was responsible for monitoring the pandemic and providing updates to global stakeholders. Very quickly, we learned that frontline responders from Manila to Milan needed real time information in their own language.

These were complex projects, with negotiations including commercial teams, compliance, engineering and delivery teams across two organisations. We’d often find ourselves in a stalemate, due to the stickiness of bureaucracy; it’s hard to shake-off old rules and processes, even in a time of crisis. It struck me that only the only people not present in the (virtual) rooms were the actual users of the end product — healthcare workers, from Manila to Milan. I found that when we were able to make progress, it’s because we tried to work backward from their needs.

Apolitical: What does it take for governments to embrace this user-centred mindset in procurement?

Quattrucci: I think it's about starting from the assumption that adopting technology is not an inevitable process — it's a skill that needs to evolve with the technologies themselves. In many public institutions, procurement is still not professionalised. It is often relegated to a compliance function, administering absurdity with no glory. But that is wrong — they should be treated as strategic negotiators. Ultimately, they're the ones responsible for upgrading the capacity of governments. Therefore if we don't invest in their skills and careers, governments won't have the capacity to absorb innovation and inventions. Balancing innovation with the demands of delivery Samasuwo: How can commercial leaders in government weigh long-term innovation against short-term cost or delivery pressures?

Quattrucci: I think we need to see commercial functions as a portfolio rather than a monolith. Within the same department, different teams should have distinct rules and permissions depending on the maturity of the market they are managing.

You can’t purchase chairs — or technologies that have been around for decades — in the same way you purchase cloud computing, AI, or quantum hardware. Yet procurement rules are often applied uniformly. Instead of framing the trade-off as short-term versus long-term, it may be better to think about diversification: some processes should be stable and predictable, while others should allow for experimentation.

That’s why you need “safe spaces” for innovation, where less mature technologies can be tested under different procurement models. At the same time, routine purchases can run through standard procedures. Agencies like DARPA show how this can work — using mechanisms like the Other Transaction Authority (OTA) to create exceptions and foster innovation.

Apolitical: Are there frameworks governments can use to embed this approach?

Quattrucci: First there are some principles public institutions can almost universally follow. For instance, the maturity of the technology should match the procurement model. I tend to think in two-by-twos: on one axis you have the maturity of the market, on the other, the procurement model. The more bespoke the product, the more bespoke the institutional model needs to be. When governments buy the things they have always bought — office supplies, for example — you don’t really need exceptions.

The should be standard processes for static markets and adaptive models for dynamic ones.

Once you look at things that way, you start to see different institutional inventions. India is an interesting case: they created procurement as infrastructure. Public digital infrastructure becomes the foundational layer of government technology — and on top of that, providers can build payments, digital ID, and vaccination certificates.

Other times you see brokering systems. Suppose multiple agencies want to procure cloud services — you have a broker administering the marketplace for them. That’s happened in the UK, in Australia, and elsewhere.

But the principle is the same: institutions need to be purpose-built tools, not blunt instruments. Otherwise, what you get is a vending machine. The way you procure a chair ends up being the way you procure cloud services — and that just doesn’t work. Adapting procurement for the AI era Samasuwo: What opportunities does AI bring for public procurement?

Quattrucci: Some companies are already using AI to simplify the initial stages of a negotiation. Imagine two parties meeting: AI can map a menu of options, and then it’s up to the humans to decide what makes sense. Machines can quickly synthesise, sort, and score options against a set of values and criteria. They can help with inventory, or generate reports on usage. If you have a bunch of compliance reports from tenders, AI can extract insights pretty quickly. There are a host of applications. But humans have the context and the responsibility to decide what to do. Something I find interesting is that the rules that governments create to purchase innovation are often so complex they scare innovators away– they run against the very intention they are set to achieve. They’re meant to encourage competition, but in practice, they become counterproductive — sometimes even creating monopolies – because only incumbents can afford to deal with slow, somewhat absurd processes. AI might also offer a way forward in this area — first to catalogue and standardise requirements, and then to automate procedures.

Apolitical: Are there changes public procurement teams need to embrace to enable this change?

Quattrucci: I like to think about commercial functions as evolving into analytics functions. One of the things that we never do or don't do often enough in government is to measure the returns on investment and that's a sort of fallacy of traditional procurement functions. We measure inputs – how much of the budget have we spent – rather than outputs or, even better, outcomes. I call it the vending machine policy, because you click your code, you get your goods and then you forget about it. But who's actually using it? If you buy 100 AI licenses, how many are being used? And for what? How? The inventiveness of an institutions stems from its capacity to learn.

Apolitical: What are some of the important capabilities that you think public servants will need?

Quattrucci: It’s important for governments —and for everyone in government — to see AI as a relational technology. What I mean is this: you’re not programming functions that a piece of software executes. You’re having a conversation with a tool that gives you answers — sometimes surprising answers, even if they’re ultimately based on statistics. The experience feels relational. It’s more like talking to a weird colleague than writing code. And that, in a way, exposes a shortage of management capacity. As tools become more sophisticated, as machines evolve into agents, using them effectively is less about technical expertise and more about communication: asking good questions, giving clear instructions, managing that weird colleague. So the slowness of organisations — not just in government — to adopt and adapt to these technologies isn’t really a tech problem; it’s a management problem. We need better managers. And AI shows that management should be treated as a real competency, not just a title.

Apolitical: That’s such a great insight. Thanks for your time Leo!

Read Leonardo Quattrucci’s paper on ‘A Well-Architected Framework for public procurement’ and follow him on LinkedIn. You can also read his opinion piece on AI adoption as a skill for governments.

Learn more about the role of AI in procurement through Apolitical’s new course on the Government AI Campus.

Cover Photo by panumas nikhomkhai.