Everyone has an opinion about what’s next for development after USAID was dismantled and European governments began cutting their aid budgets. Some say it’s time for recipient governments to invest in their own basic services and economic growth rather than relying on foreign assistance. Others are optimistic that private foundations will fill the USD70 billion gap left by USAID.
A recent article in the Financial Times argued that the new kid on the development block [chain] is artificial intelligence. AI itself will replace foreign aid. Apparently, companies like Google, Nvidia and Open AI already are delivering health, education and agricultural services once provided by development agencies.
Delusional definitions – foreign aid or forced upgrade?
Calling this “foreign aid” is like calling a chatbot your primary care provider. It confuses a service with the slow, messy business of building systems that make services stick. Aid isn’t – or shouldn’t be – about just dropping in tools, no matter how intelligent they are.
Techno-solutions or silver bullets?
Development déjà vu, anyone? Once upon a time it was microfinance, then mobile money everything, then blockchain. Now AI. The hype changes but the habit of ignoring the institutional plumbing stays the same.
The article’s examples – neonatal diagnostics, AI tutors, AI-powered tractors – are presented as stand-alone fixes, glossing over real-world delivery systems into which they must integrate. AI health tools mean little if clinics lack electricity, cold chains for vaccines, reliable transport or trained staff to interpret AI-generated outputs. AI tutors won’t help children who are hungry, absent or unconnected. Even the smartest tractor is useless without the right inputs, finance, access to markets or climate resilient infrastructure.
The add-technology-and-stir approach has rarely worked in development. It’s naïve to think that replacing GPs with GPUs will alter that. Aid – both conventional and new – has bet heavily on such silver bullets: appealingly simple solutions to complex problems, packaged as revolutionary or transformational (see also direct cash transfers and impact investing). Silver bulletism has oversold development, creating unrealistic expectations and leaving it vulnerable to attack – including, ironically, from the ultimate silver bulletists, the tech bros.
Regulatory reality check – Bretton Woulds?
The author flags the risks of dependency and value transfer in exported AI models but doesn’t dig into real-world cases of misuse, algorithmic bias or governance failures. Western governments are already struggling with AI governance, in the face of opposition from the AI companies – and we want to start exporting this stuff?
His solution is a Bretton Woods-like institution for AI governance. But in a world where support for such global bodies is declining, such noble aspirations feel like wishful thinking. Bretton Woulds, if you will.
AI can aid – but it’s not aid
The author is right that the geopolitical landscape is shifting and aid will change with it. But “code aid” risks being just another band aid unless we see it for what it is: a multiplier not a substitute.
AI deployment will rely on the very institutions it claims to bypass: regulators, utilities, telecoms, physical infrastructure and platforms for service delivery or transactions. If these institutions are under-resourced or politically compromised, then AI risks becoming yet another donor-driven pilot: technically impressive, operationally irrelevant.
How can AI aid aid?
- Recognise techno-delusion: Learn from past hype about silver bullets.
- Diagnose systems first: Infrastructure, incentives and institutional capacity before tech (function before form).
- AI as enabler, not endgame: Complement, don’t replace, human systems.
- Build absorptive capacity: Invest in institutions’ ability to manage, adapt and govern AI, not just parachute tech in.
- Guardrails and governance: Control, accountability and data rights must be part of the package, not just an optimistic postscript.
So, before we all genuflect at the altar of AI, remember: AI isn’t foreign aid, it’s a flashy sidekick. Treating AI as aid is like putting a jet engine on a donkey cart: it makes noise, it looks impressive but it’s still dragging the same load over the same ruts.
If we continue to mistake tools for transformation, we’ll just keep rebooting the same delusions with shinier branding. AI can aid aid, sure — but only if it knows its place: behind the scenes, supporting systems, not pretending to be one.