What I Told the UN About AI and Infrastructure

Last week, I joined a panel at the United Nations Economic Commission for Europe’s 9th International PPP Forum in Belgrade, focused on a topic that feels both timely and overdue: Artificial Intelligence (AI) and Public-Private Partnerships (PPPs): Potential and Limitations for the SDGs.

The session was part of a larger UNECE initiative, rooted in a growing recognition that while infrastructure investment remains critical to the SDGs, the systems we use to deliver that infrastructure—especially via PPPs—are too often slow, expensive, and out of step with what today’s challenges demand. AI, if applied thoughtfully, may be the tool that shifts that dynamic. But the path forward isn’t automatic. It’s contested, complex, and deeply human.

The panel was moderated by Stacy Sinclair, Partner at Fenwick Elliott LLP. It featured contributions from across sectors and disciplines: Syed Zaidi, Chairperson of the UNECE Working Party on PPPs; Anna Abramova, Director of the Artificial Intelligence Centre at MGIMO University; Nasser Massoud, Managing Director of Concept Realisation; David Smallbone, Partner at KPMG; Christopher Clement-Davies, Principal at CCD; and myself, representing Silta Finance.

Each of us was asked to respond to a set of guiding questions:

1. What are the most promising opportunities for AI to accelerate progress toward the SDGs thought PPPs?
2. What are some key challenges and barriers to AI adoption in PPPs, and how can they be addressed?
3. What practical steps should governments take to build the capacity, frameworks, and partnerships necessary to effectively implement AI in their PPP projects?

My task was to speak to the first—opportunities—while drawing out reflections on the others where relevant. Here’s what I shared.

The Opportunities

AI is beginning to reshape how we prepare and finance PPPs. If applied with care and focus, it has the potential to become a powerful accelerator—not just for individual projects, but for entire infrastructure ecosystems. Because the reality is this: traditional methods are simply too slow.

In the past few months alone, I’ve had conversations with PPP units, development banks, and private sector sponsors across multiple geographies. When I ask them what it would take to accelerate progress toward the SDGs through PPPs, the same two themes come up again and again:

1. We need to cut the time and cost it takes to prepare bankable projects.

2. We need to cut the time and cost it takes to conduct robust due diligence and safeguard assessments.

These are two of the most document-heavy, labour-intensive, and expensive stages in any infrastructure transaction. And too often, they’re where good projects stall—or die entirely. Especially the smaller, climate-linked projects that can’t justify high transaction costs.

This is where AI is already starting to shift the landscape.

Imagine a PPP centre—say in the Philippines—developing its project pipeline for the year ahead. Part of that process involves producing early-stage business cases or pre-feasibility studies. Traditionally, that means hiring consultants, drafting reports, conducting analysis—time-consuming, costly, and rarely consistent.

Now imagine instead that this PPP unit inputs a few key details: project type (solar), location (Palawan), estimated size (50MW). An AI tool then pulls from past project data, local energy policies, financing precedents, contractor directories, regulatory databases—producing a draft pre-feasibility study within minutes. One that outlines viable sites, regulatory considerations, cost and revenue benchmarks, possible delivery models, and risk factors.

That’s not a far-off future. It’s technically possible today. - See end of this blog for more information!

And it doesn’t stop there. Think about what happens when an unsolicited proposal lands on a government desk, and officials have just 14 days to respond. With today’s resources, that’s barely enough time to read the thing. But with AI, that same proposal could be uploaded into a platform that cross-references PPP guidelines, benchmarks costs, checks compliance, assesses ESG alignment, and produces a structured review—all in the time it takes to make a coffee.

In that moment, the PPP unit is no longer a bottleneck. It becomes a strategic engine. One that can say “yes,” “no,” or “tell me more”—based on real insight, not guesswork or political instinct.

But I didn’t want to overstate the case. AI doesn’t come without limitations.

It still makes mistakes. BUT So do humans. But AI makes them fast—which means we have time to catch and correct those mistakes earlier in the process. We should think of it as a lightning-fast assistant, not a replacement for expert judgement.

It also doesn’t know your rivers or your roads. It can’t do site visits. It doesn’t understand local politics or informal incentives. But it can take the heavy lifting off your team’s desk—freeing them to focus on high-value decisions.

And then there’s data. Governments are rightly cautious about how internal reports and documents might be used to train external AI models. But not all AI needs access to your data to be useful. There are secure, pre-trained models that can run effectively without ever compromising your control or confidentiality.

Finally, there’s the human element. Every organisation has early adopters and late adopters. If we want this shift to stick, we need not just training, but governance and incentive alignment. People need to understand what these tools are, what they aren’t, and how to interpret the outputs. Because AI is a precision tool—not a universal fix. Just because we’ve found a hammer doesn’t mean everything is a nail.

So what should governments do?

I offered three practical steps.

First, invest in clear data governance. Define what can be used, who owns it, and how it’s safeguarded. - Importantly, this should be done swiftly, i.e within weeks, not months.

Second, create sandboxes. Run controlled pilots inside PPP units on real, document-heavy workflows—without high stakes. Build institutional muscle memory before attempting system-wide change.

Third, train your people. And not just technical staff. Everyone. Because the success of AI in PPPs will depend less on algorithms and more on understanding. The best models in the world won’t help if the people using them don’t trust them—or can’t explain them.

I closed my contribution with a simple truth: AI won’t fix broken institutions. But it can reveal the cracks faster. And that gives us time to act.

If we get the policy right, if we build the right frameworks and the right teams, AI has the potential to dramatically speed up how we develop, finance, and deliver infrastructure—making PPPs more inclusive, more efficient, and more aligned with the goals we’ve all signed up for.

It’s not a silver bullet. But in the right hands, with the right intention, it might just be the accelerant we need.

From Ideas to Action: What We’re Building at Silta Finance

At Silta Finance, we’re not just exploring these ideas—we’re already implementing them.

Over the past 14 months, our team has worked closely with a multilateral development bank to deliver faster, more consistent due diligence and ESS assessments using Silta AI. What began as a proof of concept is now a proven platform—battle-tested across a variety of infrastructure projects throughout Southeast Asia, Central Asia, and South Asia.

We’ve seen the results firsthand. On average, Silta AI is already delivering time savings of 40% or more across key project workflows. In some cases—particularly for repetitive document analysis and safeguard reviews—the time savings are even higher. And this is just the beginning.

Six months from now, we expect that number to approach 80%. Not because we’re cutting corners, but because we’re removing the bottlenecks that slow down human decision-making. This doesn’t eliminate jobs—it expands capacity. The time saved becomes time re-invested. Time that governments and development banks can use to assess more projects, respond to more proposals, and move good ideas to funding faster.

We’re now beginning to support governments directly—helping them evaluate competitive tenders using AI-assisted assessments aligned to their own procurement guidelines and strategic priorities. The goal isn’t to replace public decision-making. It’s to enhance it—to give teams the tools they need to make faster, more transparent, more consistent decisions.

Silta AI has been built by people who understand infrastructure finance. That’s why it works. And that’s why we’re so confident in what comes next.

If you’re a PPP unit, a multilateral agency, or a project sponsor trying to do more with less—and do it better—now’s the time to talk.

Because the future of PPPs isn’t just faster. It’s fairer, smarter, and more accessible. And we’re ready to help you build it.

If you’re navigating your AI strategy in infrastructure—whether you’re in government, a development bank, or advising PPP projects—and you’re unsure where to start or how to scale, I’d be happy to talk.

And if you’d like to see Silta AI in action—how it’s already supporting due diligence, ESG assessments, and tender evaluations across Southeast Asia, Central Asia, and South Asia—just drop me a message.

This isn’t theory. It’s working today. Let’s explore what it could do for you.