AI and the IKEAfication of Impact

The economics of good enough, the disappearance of craft, and what happens when producing development becomes easier than understanding it.

IKEA understood something that much of the world took decades to appreciate. Most people don't need exceptional furniture. They need furniture that looks good, works reasonably well, is affordable and can be assembled without calling a carpenter.

It was a brilliant proposition. IKEA didn't simply make furniture cheaper. It redesigned the economics of an entire industry around the principle of just good enough.

And it worked.

Today, something remarkably similar is happening in the development industry. Artificial intelligence is accelerating a transition that was already well underway: from bespoke interventions to standardised products, from professional judgement to prescribed methodologies, and from craftsmanship to assembly.

This is not necessarily a bad thing. In fact, much of it is extraordinarily useful.

But it has consequences.

The economics of good enough

For decades, development organisations have invested heavily in the production of intellectual infrastructure. Strategies, programme designs, theories of change, implementation manuals, monitoring frameworks, funding proposals and evaluation reports.

These products were expensive partly because producing them required people with specialised knowledge and experience.

AI has fundamentally altered that equation.

A reasonably competent user can now produce, in an afternoon, documents that previously required weeks of professional effort. The formatting is excellent. The terminology is appropriate. The logic is generally coherent. The indicators are measurable.

And increasingly, the result is good enough.

This matters because much of the development economy doesn't purchase excellence. It purchases confidence that an intervention is credible, compliant, fundable and manageable.

AI is becoming exceptionally good at producing that confidence.

The disappearance of the craftsman

The IKEA analogy becomes more interesting when we consider what happened to furniture making.

IKEA didn't eliminate good furniture design. It concentrated much of the expertise in the design and manufacturing system while making the final product accessible to people who possessed none of that expertise.

But the market for traditional craftsmanship changed.

The person who understood timber, joints, materials and construction was no longer necessary for most furniture purchases.

In development, we're witnessing a comparable redistribution of expertise.

The practitioner who knows how to negotiate competing community interests, recognise institutional dysfunction, design around political constraints or understand why a theoretically sound intervention will fail in practice is our equivalent of the craftsman.

That knowledge is difficult to standardise. It is accumulated through experience, failure, relationships and repeated exposure to complexity.

Yet increasingly, the products of that expertise can be simulated without the experience that originally produced them.

And because AI-generated work is often indistinguishable from professionally produced work at first glance, the market may struggle to recognise the difference.

The apprenticeship problem

Perhaps the most significant consequence is not the displacement of experienced practitioners. It is the weakening of the process through which new practitioners become experienced.

Craftsmanship is learned through doing.

You become a good programme designer by designing programmes, getting things wrong, confronting unexpected realities and learning to distinguish between what looks convincing and what actually works.

You become a good strategist by wrestling with ambiguity, making decisions with incomplete information and living with their consequences.

AI can now remove much of that struggle.

The junior practitioner who once spent three days constructing a programme framework can generate one in three minutes. The analyst who once had to understand the relationship between evidence and recommendation can ask AI to produce both.

Productivity improves immediately.

But what happens to the learning that occurred during those three days?

We should be careful not to romanticise inefficient work. Much of it was tedious and added little educational value. AI can also accelerate learning when used intelligently.

Nevertheless, there is a material difference between using technology to develop judgement and using it to avoid exercising judgement altogether.

We are making it easier to produce the work while potentially making it harder to develop the people who understand the work.

The industrialisation of impact

The development sector was already moving towards an IKEA model before generative AI arrived.

Standardised programme packages. Replicable interventions. Common measurement frameworks. Centralised programme design with decentralised implementation. Increasing emphasis on scale, efficiency and cost per beneficiary.

These developments have considerable merit.

There is no reason every organisation should independently design an attendance register, develop a safeguarding policy or invent its own monitoring architecture. Standardisation can release resources for activities that genuinely require contextual judgement.

The problem arises when we begin to standardise things that are not actually interchangeable.

Communities are not identical. Institutions behave differently. Local economies have distinct incentives. Political relationships matter. The same intervention can produce very different outcomes depending on who delivers it, where, and under what conditions.

IKEA can manufacture a bookshelf to precise specifications because the physical properties of the product are relatively predictable.

Social change is less accommodating.

And AI, trained to recognise and reproduce patterns, can make those differences appear less consequential than they are.

The market may not notice

Here is where the economics become particularly uncomfortable.

If a funder cannot readily distinguish between an intervention designed through deep contextual engagement and one assembled from existing templates, why would they pay substantially more for the former?

If both proposals contain credible objectives, coherent methodologies, attractive budgets and convincing indicators, what exactly is being purchased?

The market may increasingly reward the ability to produce credible development products rather than the ability to develop credible solutions.

That doesn't mean funders are foolish. It means the signals traditionally used to assess professional competence are becoming less reliable.

A beautifully structured strategy once suggested that considerable intellectual effort had gone into its development.

Today, it may simply indicate that someone knows how to write a good prompt.

The distinction only becomes apparent when implementation encounters reality.

And by then, the proposal has been approved, the funding committed and the reporting framework established.

The craftsman with a power tool

There is, however, another possible outcome. AI could make genuine expertise substantially more valuable.

An experienced practitioner who understands institutions, incentives, politics and implementation can use AI to eliminate enormous amounts of administrative and analytical work.

They can test scenarios faster, interrogate assumptions more thoroughly, compare alternative approaches and produce sophisticated outputs without maintaining a large supporting bureaucracy.

In that model, AI doesn't replace craftsmanship. It increases the productive capacity of the craftsman.

The competitive advantage shifts away from the ability to produce professional-looking documents and towards the ability to ask better questions, recognise weak assumptions, exercise judgement and navigate reality.

This could favour smaller, highly capable organisations over larger institutions whose advantage historically depended on administrative capacity and production infrastructure.

But that outcome is not guaranteed.

Markets do not automatically reward deeper expertise. They reward what purchasers can recognise, value and afford.

If the purchasing system continues to privilege standardised outputs, genuine craftsmanship may become economically marginal even as its practical importance increases.

What happens when everyone can assemble?

IKEA's achievement was to make functional, attractive furniture accessible to millions of people who could never have afforded bespoke craftsmanship.

AI offers a comparable opportunity in development.

Smaller organisations can access analytical and production capabilities previously reserved for well-resourced institutions. Local practitioners can develop sophisticated proposals without expensive consultants. Community organisations can participate in funding processes that historically excluded them.

That is a meaningful redistribution of capability, at least at the level of production.

But access to the tools of expertise is not the same as possession of expertise.

And this is where the next phase of the development economy becomes interesting.

We may see an abundance of competent-looking programmes, an expanding market for standardised interventions and a significant reduction in the cost of producing the administrative machinery of development.

At the same time, the people capable of designing genuinely original responses to complex social problems may become relatively scarce.

Not because AI has made them unnecessary, but because the economic incentives to develop their skills have weakened.

The industry could become extremely efficient at reproducing what it already knows while gradually losing some of its capacity to discover what it doesn't.

Beyond the flat pack

The IKEAfication of impact is not a warning against standardisation, scale or artificial intelligence.

IKEA's proposition is brilliant precisely because it recognises that not everything needs to be exceptional. There is enormous economic and social value in making things just good enough, consistently and affordably.

Development should learn from that.

But it should also recognise the difference between manufacturing a product and changing a system.

A flat-pack bookshelf does not need to understand the room in which it stands.

A social intervention does.

The opportunity is to standardise the machinery of development without standardising away the intelligence required to make it work.

AI can make that possible. It can also make the opposite dangerously easy.

We are entering an era in which almost anyone can assemble an impact programme. The question is whether we are still producing people who know how to build one.

And perhaps, more importantly, whether anyone will still be willing to pay for them.

This is how we #GrowZA