What George Osborne’s New Job Tells the UK Aid Sector About Its Future

George Osborne, former Chancellor of the Exchequer, architect of austerity, occasional newspaper editor, is now four months into a new role as Managing Director and Head of OpenAI’s “for countries” programme.  

News in recent weeks that Osbourne will be working out of a new 88, 000 square foot office in King’s Cross, with 544 desks, gives a sense of the scale and ambition of this operation.

Whatever you think of Osborne the politician, the symbolism here is hard to miss. One of the most consequential figures in a generation of British public life has placed his bet firmly in the AI-for-development space. And he’s not doing it through the FCDO, or the World Bank, or a think tank. He’s doing it from London and through OpenAI, the most powerful private AI company on earth, in a role and mission that’s explicitly backed by the full weight of the United States government.

The UK aid sector would do well to pay attention. Not to Osborne per se, but to what his appointment, and the size, scale and ambition of the programme he now leads, tells us about where the world is going. 

The New Architecture of Global Problem-Solving

OpenAI for Countries is not foreign aid. Let’s be clear about that. It’s commercial, transactional and unapologetically designed to extend American technological and geopolitical influence. As the White House’s AI Action Plan frames it, the goal is nothing less than “removing barriers to American leadership in AI” with the Commerce and State Departments partnering with industry to deliver what they call “secure, full-stack AI export packages” to friends and allies around the world.

Hardware. Models. Software. Applications. Standards. The whole stack. Deployed globally. Backed by the US government. With George and Sam Altman’s people doing the deals.

This is the privatisation of the kind of institution-building that aid used to do. The capacity building, the systems reform, the public service transformation, all of it now repackaged as a commercial proposition, with Altman’s stated aim of helping countries build “national AI infrastructure” that will be “the backbone of future economic growth and national development.”

The FT called it well last summer: AI is the new foreign aid. OpenAI for Countries is the proof of concept.

The countries on the receiving end of this package aren’t being offered a choice between the old model and the new one. They’re just being offered the new one. The bilateral donors have retreated. The US has hollowed out USAID. The UK has cut ODA to less than 0.3% of GNI. The multilaterals are being restructured, with much of their new contracts delayed until the autumn. The vacuum is being filled rapidly, commercially, at scale by tech.

How Fast This Is Moving

When I first wrote on OpenAI for countries in February, Osborne was barely a month into the job. The programme was still largely promissory, a handful of announced partnerships and a lot of rhetoric about democratic AI rails. What has happened since should concentrate minds across the sector.

Osborne’s formal title is Managing Director and Head of OpenAI for Countries, based in London. He started in January 2026 and was immediately deployed to the World Economic Forum at Davos, where the programme announced a major expansion. OpenAI has now engaged with more than fifty countries. Over thirty governments have expressed interest in formal partnership. Announced deals already span Argentina, Australia, Germany, Greece, Ireland, Norway, the UAE, South Korea, the UK, and Nigeria, with pilot infrastructure sites in Nigeria and the UAE expected to have gone live by now.

But the really significant development is scope. This is no longer just an infrastructure play about Stargate data centres and compute capacity, important as that is. At Davos, Osborne announced new programme streams for 2026 covering education, health, AI skills training and certification, disaster response and preparedness, cybersecurity, and startup accelerators. There is now a dedicated Education for Countries sub-programme: Estonia has already deployed ChatGPT Edu nationwide across universities and secondary schools, reaching over 30,000 students and educators, with a longitudinal research partnership with the University of Tartu and Stanford measuring learning outcomes among 20,000 students over time.

In Europe, OpenAI has launched “OpenAI for Europe” as a regional adaptation, complete with an SME AI Accelerator in partnership with Booking.com targeting 20,000 small businesses. In Japan, there’s a strategic collaboration with the Digital Agency aligned to the Hiroshima AI Process framework. The programme has its own intellectual apparatus too, Osborne’s signature concept is the “capability overhang,” the gap between what frontier AI systems can do and how countries are actually deploying them.

His research claims power users rely on roughly seven times more advanced thinking capabilities than typical users. That’s the pitch to government ministers: you’re not falling behind because you lack access; you’re falling behind because you’re not using what’s already in front of you.

All of this in four months. With a former Chancellor running it out of London.

The Knowledge Is Already In The Room

Here’s the genuinely exciting part.  The UK aid sector (battered, bruised, shrinking) sits on an extraordinary and largely unrecognised asset. Two decades of hard-won knowledge about how to do things in difficult places. How governments under pressure, emerging from conflict or political transition actually work, and don’t work. How to navigate the gap between policy intent and institutional reality. How to build reform coalitions in environments where reform has powerful enemies. How to manage fiduciary risk in fragile states. How to engage with civil society and local actors in ways that actually deliver change.

This is Thinking and Working Politically. This is adaptive management. This is political economy analysis. This is the stuff that OpenAI for Countries will need (desperately) if it is going to do more than install infrastructure and leave. Because the history of technology-enabled development is littered with platforms that worked perfectly and changed nothing, because nobody did the hard political and institutional groundwork to make change stick.

The challenge for the UK aid sector is to stop seeing this knowledge as the thing that used to be valuable, and start seeing it as the thing that becomes more valuable as the technology scales.

Governments in the Global South are not short of AI products being pitched at them. They are short of trusted advisers who understand their political economy, who can help them navigate the pressures of the US-China technology competition, who can help them think clearly about what national AI infrastructure actually means for their citizens and their civil service and their democratic institutions.

That’s the expertise that’s locked up and beginning to languish in the UK aid sector.

And it’s worth noting what’s conspicuously absent from everything Osborne and OpenAI are saying. There’s plenty about infrastructure, compute, education tools, skills certification, startup ecosystems. There is nothing, not a word, about the political economy of technology adoption. Nothing about how you actually get a ministry of finance that doesn’t trust its own planning commission to agree on a shared data architecture. Nothing about how you build reform coalitions in governments where the people who benefit from opacity are the ones controlling procurement. Nothing about what happens when AI-enabled transparency tools, run headlong into the interests of entrenched elites. (MetricsLed-DECLARE is an asset declaration and unexplained wealth tracking solution).

It seems to this veteran of the sector that this silence is a market signal. It’s the gap between what OpenAI can sell and what countries actually need. Somebody has to fill it.

From Delivery to Deployment

The pivot I’m describing is not as dramatic as it sounds, but it does require an honest reckoning with the direction of travel.

The era of the large UK-headquartered prime contractor running a five-year bilateral project with a team of expat advisers and a logframe the length of a small novel is drawing to a close. But what replaces it is not a void. It’s a new and rapidly expanding market in which governments around the world need help working out how to harness AI for public service delivery, for revenue mobilisation, for asset transparency, for health systems, for education, for the kinds of governance reform that donors used to fund through traditional TA.

OpenAI can sell them the infrastructure. Who helps them use it? Who does the capacity building? Who advises on the institutional reform needed to make a national AI strategy more than a PowerPoint? Who runs the political economy analysis that identifies where AI-enabled service delivery will face resistance and why? Who helps them design an AI-ready civil service?

These are not questions that Silicon Valley is well placed to answer. They are questions that people who’ve spent careers navigating the realities of development, who know what it’s like to try to reform a public financial management system in a country where the finance ministry and the presidency aren’t on speaking terms, are exceptionally well placed to answer.

The opportunity is real. But it requires a decisive pivot.   Firms who have seen themselves as aid delivery organisations must recast themselves as specialist advisory firms with deep expertise in complex institutional environments, now operating in a world where the tools available to address those challenges are orders of magnitude more powerful than they were five years ago.

Turning Ploughshares into Algorithms

The White House talks about AI ushering in a “new golden age of human flourishing.” It’s the kind of language that makes development professionals reach for the smelling salts. We’ve heard ‘the latest big thing’ in aid narrative before, applied to structural adjustment, to the Millennium Development Goals, to big data, to making markets work for the poor (M4P) to unconditional cash transfer.   The cycle of promise, of lemming-like rush into the new big thing, through to overreach disappointment, wind down, and report shelving. 

But this time it’s different in one important respect: the technology actually works. AI systems are already diagnosing disease in rural clinics, predicting crop failures, identifying fraud in public procurement systems, translating government services into languages that were previously unserved, and doing it at a fraction of the cost and a multiple of the speed of any previous intervention.

The question for the UK aid sector is not whether this wave is coming. It is already here.

George Osborne’s new job is a signal flare. OpenAI’s “for countries” programme is a signal flare. The White House AI Action Plan’s international dimensions are a signal flare. The FT writing about AI as the new foreign aid is a signal flare.

And the speed at which the programme has scaled since January, fifty countries engaged, dedicated verticals in education and health, a European regional strategy, a conceptual framework being road-tested at Davos and Delhi, tells you that this is not a pilot. This is an operating model. It is being built at pace, by very capable people, with very deep pockets, and it is filling the space that the UK aid sector and all the other traditional aid agencies are vacating.

The sector that adapts (takes its political economy expertise, its stakeholder engagement skills, its experience in public sector reform, and wraps it around the new tools and the new delivery models) will find that there is significant and growing demand for what it knows how to do.

A Practical Start

So what does pivoting actually look like? A few concrete thoughts:

The first move is to invest in understanding the tools. Not at a conceptual level, at a hands-on level. The organisations that will thrive in this space are those whose senior advisers understand what AI can and cannot do in a government reform context. That requires investment in training, in experimentation, in building internal AI capability.

The second is to rethink your offer. The ‘Technical Assistance’ model still works, but the technical assistance that is valuable has changed. Helping a Ministry of Finance implement an AI-powered tax compliance system requires the same political economy analysis and stakeholder engagement that helping them implement any other reform requires, but it also requires knowing enough about the technology to be a credible interlocutor. Build those teams now.

The third is to follow the money – the new money. The US government’s AI diplomacy push, OpenAI for Countries, the Gulf’s investment in digital infrastructure, the growing number of multilateral funds focused on AI for development — these are the emerging procurement pipelines. The FCDO pipeline will shrink and change. The new pipeline is already taking shape. And OpenAI’s own expansion — the $500 billion Stargate project, now exceeding its initial targets with over 5 gigawatts of data centre capacity under development — tells you the capital behind this is not speculative. It is committed and deploying.

Fourth, is to resist the temptation to treat AI as a cost-cutting tool and nothing more. Automation matters, and yes, it should be deployed to drive down overhead and sharpen competitiveness. But the much larger prize is repositioning AI as the subject matter of your advisory offer, not just the means by which you deliver it. That’s where the growth is.

The fifth, and perhaps the most important, is to reach out to experts on AI with real world data collection, data management, and cybersecurity expertise.  Experts from within our sector to help capitalise on AI’s capability and use the insights it will bring to support aid and stabilisation. One number on that speed dial should be…..  MetricsLed™.

Conclusion

George Osborne in Delhi, telling developing country leaders not to be left behind, will be rich with irony for some of those who remember his tenure at the Treasury. But the message itself is not wrong. And it applies, with equal force, to the UK aid sector that his government did so much to shape and that his successors have done so much to diminish.

Don’t be left behind.

The knowledge is there. The expertise is there. The need is there, and growing. The tools are more powerful than anything the sector has previously had at its disposal. And the pace at which the alternative model is scaling (OpenAI for Countries went from announcement to fifty-country engagement in under a year) should tell you that the window for positioning is not indefinite.

What’s needed now is the courage to pivot – decisively, intelligently, and soon – from delivering the old model of aid to advising on the new model of AI-enabled development.