Biology is an elegant operating system. It has been driving the machinery of living organisms for a long time and doing so seemingly autonomously. Rather than using bits, information is represented by atoms, and it can execute programmes written in an obscure yet highly performant language known as DNA. The execution of unique source code within each living thing gives rise to the infinite diversity in their structure and function.

Human health is predicated upon understanding the inner workings of this operating system — how it performs instructions, handles I/O, and maintains state. The challenge has always been that it resembles a black box — incredibly complicated and has no documentation. Since the early days of civilisation, humans have sought to decode it by way of observation and experimentation, one unit of endeavour at a time. Recording of findings and more systematic investigation led to continuous accumulation of a shared knowledge over time about its underlying intricacies, which today forms the foundation of modern medicine. Despite remarkable achievements in healthcare owed to this increased understanding, there is much left to discover about the system.

Augmented intelligence — 20 experts for the price of 1

The complexity of biology and the need for various human experts to interpret, explore, perturb, and consult on it are core drivers of what makes healthcare expensive today — often prohibitively so. New tools and technology have advanced the field — allowing for reading source codes, interpreting stack traces, writing new programmes, and analysing output — and produced more data than could ever be obtained before. Yet, a tangible reduction in cost of care for patients remains elusive.

This, however, may be changing. The confluence of deep learning models, powerful computers, cloud infrastructure, and data has put artificial intelligence front and centre in the evolution of several industries — of which healthcare is one of the most promising. By supercharging human efforts, AI has the potential to radically optimise cost across the healthcare value chain — particularly in medical research, drug development, and patient care.

Weaving a tapestry of biology, one petaflop at a time

AI has the potential to accelerate scientific discovery in medicine, saving hours of work and expensive resources.

It can aggregate existing knowledge about biology and piece together a more thorough blueprint of its underlying mechanics — akin to a form of documentation. By interacting with this documentation, scientists can more quickly extrapolate new knowledge — e.g. deciphering the semantics of biology’s programming language, elucidating disease origin and mechanism of action, discovering novel biomarkers, and identifying areas for further work in the wet lab.

These AI systems can be built by training deep learning models on diverse biological datasets, including omics data, clinical data, and scientific literature, with the potential for specialisation or generality and the ability to process uni- or multi-modal information.

A notable example of such models is DeepMind’s AlphaFold. Trained on a proteomics dataset, AlphaFold learned to predict the 3D conformation that a newly manufactured protein comes to adopt within the body of a living thing — critical for understanding disease and how to treat it. It can do this for any of over 200 million known proteins within minutes and with a high degree of accuracy. Structures had been typically studied manually and could take a researcher an entire PhD programme to determine for a single protein.

Therapeutics 2.0 — from empiricism to engineering

AI has the potential to bring more new medicines to market, making the development process quicker and less expensive.

The drug development process is effectively a search problem — one in which the goal is finding a chemical or biological asset (often a molecule) that can successfully modulate a specific pathway within the body that is involved in a given disease, and thereby elicit a therapeutic effect that is consistent across a broad sample of patients, all while not triggering adverse side effects.

The search space is immense, and navigating it is a long, risky, multi-stage undertaking that typically involves a combination of trial and error, heuristics, and serendipity. This inflates the price of medicine and limits resource allocation to only disease areas with high economic potential.

AI can transform the process by accelerating design-build-test-learn cycles, optimising search efforts to only those likely to succeed, and shifting search workloads to the more efficient realm of silicon. Not only can this optimise process cost, but also lower the barrier to entry in terms of tackling disease and broaden the universe of those that can be tackled economically.

Seeking a fourtieth opinion?

AI has the potential to make clinical care more efficient, boosting the productivity of care providers.

It can support provider decision making, helping them extrapolate more insight from patient data — whether genotypic or phenotypic. It can synthesise and distil complex medical information (e.g. patient records), pick up on hard-to-discern patterns (e.g. in medical images), interpret raw health data in real-time (e.g. from remote sensors), and make “educated” recommendations (e.g. screening based on genetic risk factors). This can enable more accurate and timely diagnoses, treatments, and interventions.

AI can also streamline their workflow, eliminating administrative burden. It can transcribe patient interactions during consultations and serve as a repository for patient records that can be interacted with in natural language — akin to a super assistant.

Patients can, in turn, receive maximised value — more effective and personalised care, and an overall better experience.

A new day for biotech

Such intelligent capabilities will find their way into the bio and health industry as next generation software products — ones with seamless interfaces (as APIs or other UIs) and built by a new breed of entrepreneur. These new biotech founders are multilingual — fluent across data, life, and computer sciences — and of a tech-first mindset.

Indeed, new AI-native business models are emerging within biotech. Therapeutics companies, rather than being built around the development and commercialisation of a single lab-derived hypothesis or asset, are architected as data-driven discovery platforms, off which multiple assets can be engineered. Their proprietary data, which is generated at scale via high throughput biology, is their core competitive advantage versus single IP. An early wave of such companies emerged over the past decade, with currently several of their assets undergoing clinical trials.

The space is nascent and there is still a lot of ground to cover. The bio and health value chain is becoming more unbundled and traditional demarcations increasingly blurred. Advances in life science, machine learning, biotech tools, medical devices, manufacturing processes, software stacks, and open sourcing will dictate how the sector will evolve. Barriers to entry are likely to be lowered, exerting more competitive pressure on incumbents.

Move carefully and don’t break things

In contrast to mainstream tech, missteps in bio are serious. New products should be built with care, caution, and rigorous testing. While hallucination is passable in many current applications of generative AI, it is not in healthcare. Explainability is also critical, allowing machine predictions to be validated by experts. There are nuanced legal and ethical considerations at play.

There are also questions of privacy, protection, and bias when it comes to patient data, as well as questions about wellbeing — specifically biosafety and biosecurity. There is arguably more need today for talent in the global regulatory sector than ever before, to address these issues and pave the way for the technology to be deployed responsibly.

The future hasn’t been written yet — but all the ingredients are there to make this a giant leap forward for human health.

The mind has fascinated humans for a long time. Science has through time been employed to explain it and today has a dedicated branch known as cognitive neuroscience. With the advent of modern computing, a different but related experiment had arisen — replicating the mind. This is the discipline known today as artificial intelligence, or AI for short.

AI is a field broadly concerned with simulating human intelligence in machines. While dating back several decades, it has been propelled to the fore by twenty-first century advances in technology. Considerable progress has been made, with machines today being able to “see”, process natural language, make predictions, take decisions, create content, and move around in the physical environment. These abilities, however, currently vary in their depth and breadth when compared to humans — some are sharper, others not. And current AI systems are generally able to manifest only each at a time — a limitation known as narrow AI. This contrasts with general AI systems which are able to perform any intellectual task that a human can — a highly sought-after achievement by some industry practitioners.

Machine learning demystified

The remarkable thing is that most machines today are not explicitly programmed with instructions on how to perform such complex tasks, which had been the challenging modus operandi in times prior. Instead, machines “learn” how a task is performed by recognizing patterns in data — a technique called machine learning (ML). This technique has been game changing — so much in fact that some leading practitioners have hailed it as a paradigm shift in how software is engineered.

Machine learning systems are driven mainly by two elements: data and a model. In plain language, data is fed to a model as “examples”, from which the model gains “experience”, and is then able to make “guesses” from what it has learnt. Imbibing a model with experience is a stage of the process called training, while eliciting guesses from a model is one called inference. These “educated guesses” are what allow AI systems today to manifest sophisticated abilities.

Architecting an ML system involves, amongst other things, defining what task needs to be engineered, what data is required to engineer it, and selecting a suitable model to train based on the nature of task and data.

Deep learning — machine learning on steroids

Today, there is much more data that machines can learn from than ever before. Increased digitisation, broader internet penetration, as well as the availability of new platforms, apps, and devices have all contributed to this explosion. Where data is scarce, it is also generally easy to create/capture using the ample technologies that exist. Data pipelines have become easier to create, manage, and scale due to the proliferation of cloud computing options. This has allowed ML to flourish and its use cases to grow exponentially.

There have also been significant advances on the model front over the past two decades. ML models are typically computationally expensive to train — they are based on optimisation algorithms that require considerable computing resources to run on a dataset. Twenty-first century advances in computing power and access to such distributed resources on the cloud have allowed for the training of larger models on larger datasets. A particular class of model that scales very usefully this way is neural networks, which are of an architecture inspired by the structure and function of the human brain.1 Training large neural networks on big datasets is a type of ML called deep learning.2 Deep learning models have proven to be powerful — they learn with less need for human engineering and generally outperform traditional models.

Foundation models — an emerging paradigm

There has lately been an outpouring of research into deep learning within both academia and big tech, which has produced new model variants that learn in more meaningful ways — seemingly providing models with a deeper understanding of the world. Instead of the traditional single-task approach, these models can be pre-trained on large corpuses of data and then adapted to perform a wide range of downstream tasks. They can also learn from multimodal data (e.g. text, images, speech) then be adapted to tasks spanning several modes (e.g. vision and language). Adaptation typically requires additional training on domain-specific data, but a remarkable feature of some of these models is that they can also be adapted via in-context learning — a process akin to learning on-the-fly by receiving a brief explanation of a task in natural language. Such models are termed foundation models, as a reflection of their platform-like nature and status as an emerging ML paradigm.

Particularly popular applications of foundation models at the moment are conversational AI and AI art –– both forms of generative AI based on language and text-to-image models respectively.

While ML breakthroughs are still mostly confined to the research domain as well as within big tech companies (which are wealthy in data, talent, and financial resources), foundation models, tooling, and a growing open-source community contribute to its democratisation and are expected to accelerate its diffusion from research to industry in the coming years.

Artificial intelligence in practice

AI has the potential to both create considerable economic value as well as increase its propensity to change hands. This has not gone unappreciated at the geopolitical level, with what seems to be an emerging AI race between global powers. This impact will also be felt by businesses across all industries and geographies, with existing businesses adapting to a new playing field (e.g. cloud migration, being data driven, automation) and new ones are emerging because of it, especially within the last few months and years.

The industry’s increasing prominence has resulted in the term “AI” becoming a buzzword that is often used liberally, which can be confusing for many people. To better understand AI companies, it can be useful to classify them into one of three categories: Applied AI, Model-as-a-Service, and AI Infrastructure. Some companies may span several categories.

Applied AI

Applied AI companies are those that leverage AI capabilities to drive their business. These can be legacy companies like Netflix, Baidu, Tesla, or GitHub that have re-engineered existing offerings to use AI, or “AI-first” companies like PolyAI, Altana, Freenome, or Jasper that have built intelligent products from day one. While legacy companies that apply AI have mostly been internet companies, they can also be traditional businesses — e.g. a hospital that uses an ML model to diagnose conditions from medical imaging, a design office that uses a text-to-3D-object model to render mock-ups.

Applied AI companies can also be classified as vertical or horizontal, meaning they are specific to a particular industry or applicable across industries. For example, Freenome has developed an intelligent cancer detection product (i.e. healthcare), while PolyAI’s voice assistants have applications in a variety of industries (e.g. retail, hospitality, banking).

Applied AI companies maintain their competitive advantage through various strategies, such as proprietary data, proprietary models, and vendor lock-in. Newer companies can initially make use of open-source data, open-source models, and foundation models, then differentiate themselves as they build momentum.

Model-as-a-Service

Not dissimilar to Platform-as-a-Service, the past years have seen the emergence of Model-as-a-Service (MaaS) companies. They leverage in-house ML engineering teams and access to significant computing resources to develop proprietary foundation models. Such models are pre-trained on vast amounts of data from various sources, including the internet, and their usage is generally paywalled behind an API or some other UI, with inference taking place on company servers — a typically expensive incurrence. They can be used as is or adapted for specific use cases — including by applied AI companies.

While MaaS is a nascent business model, it is expected to multiply in the coming years as foundation models become more mainstream and AI adoption ramps up in industry. Examples of MaaS companies include OpenAI which develops the GPT family of language models (including ChatGPT and GPT-3), the visual language model DALL-E 2, and the text-to-code model Codex; Stability AI which develops the Stable Diffusion family of visual language models; Midjourney which develops eponymous visual language models; Google which hosts the text-to-speech model WaveNet3; and AI21 Labs which develops the language model Jurassic-1.

AI Infrastructure

AI infrastructure is a “backend” industry and encompasses a range of companies whose products support the AI process.

These can be MLOps4 products, which are a suite of software products that help humans in developing and operationalising ML systems. These products can provide workflow management, cloud resources, tooling, and versioning across the entire ML lifecycle, and are considered key catalysts to the adoption of AI in industry. There are many flavours of MLOps out there, including those that specialise in a particular stage of the lifecycle (e.g. data annotation, model fine-tuning), end-to-end solutions, tools for low-code users or experienced engineers, and tools that are specific to a single domain (e.g. vision, language) or applicable to a wide range of them. Cloud providers like Snowflake, Google, Microsoft, and AWS have moved into this space, leveraging their position as entrenched data platforms. “AI-first” companies also compete including Landing AI, Databricks, Snorkel AI, and Modular.

AI infrastructure companies also develop specialised hardware, such as AI microchips. These are designed to accelerate ML applications, as traditional general-purpose processors are inefficient at running heavy ML workloads. Companies that manufacture these microchips often adopt a cloud computing business model, and include Graphcore, Cerebras, Esperanto AI, and Google.

AI microcontrollers are another type of specialised hardware that allow edge devices to handle deep learning inference workloads without needing to send data to a remote server, improving the speed and efficiency of the process. Applications that benefit from this include those that require low-latency inference at the edge, such as autonomous driving and surveillance. Examples of companies that produce these microcontrollers are Hailo AI and Arcturus.

Inferring the future

There are certain trends in AI that are useful to note, especially regarding the future ahead. These trends include the increasing demand for computing power, the growing role of edge AI, the homogenisation of models, the potency of intelligence research, the rising importance of ethics & regulation in the field, the need for workforce reskilling, and the dangers of data concentration.

The jury is still out on whether AI will be able to match the human mind in its singular wonder. Regardless of the outcome — brace for change.

There are more than 510 million smallholder farms in the world producing some $2tn worth of food annually.1

Despite their importance in global food systems – representing 84% of all farm holdings and producing one third of all food2 – they face undue economic marginalisation and hardship.

Most smallholder farms live in poverty, insecurity, and constant need of complementing their insufficient income with off-farm streams. They also face a succession planning issue – the smallholder farmer population is aging while younger generations opt out to pursue other more prosperous occupations elsewhere, often in cities.

Farmer pain points

Two chief reasons for smallholders’ current state of affairs are market exclusion and low competitiveness. Their scale and fragmentation pose a challenge for them to access end markets for their produce which limits their agency and bargaining power. And their resource poverty leaves them unable to invest in quality inputs thereby making it difficult to compete with larger, more industrial farms.

Agro-cooperatives, which exist to alleviate such pains, are limited in reach, scope, tech intensity, and success – especially in regions with a high concentration of smallholder farming.

Beyond agro-specific challenges, smallholders are also the most vulnerable cohort to economic (e.g. COVID-19 supply chain disruptions) and climate-related (e.g. extreme weather events) shocks and stresses – events that are happening with more frequency and pose a threat to their livelihoods.

Wide lens view

In spite of this, there is a broader social, economic, as well as environmental need for a livelier smallholder farm existence.

To cope with growing demand, food systems will need to deliver 50% more food by 2050.3 The FAO estimates that, with current resource constraints, the majority of this increase will come from yield enhancement – an opportunity which exists predominantly within the smallholder farm domain.

The resilience of food systems is also a topic of increasing concern, particularly in the wake of COVID-19 disruption. The increased diversity and quantity of food source nodes, as well as of food trade flows, are characteristics of a more resilient system4 – ones which can only be achieved by better integrating smallholders into the global food supply chain.

Large-scale, intensive mono-cropping is degrading the quality and sustainability of our natural environment,5 which poses a threat to life on Earth. Smallholder farm success is one of several elements that are key to maintaining biodiversity and the sustainability of agricultural systems.

Technology in farming

The plight of smallholder farms is unjustifiable in this day, where tech has the capacity to level the playing field as well as make the occupation “sexy again” for a new tech-savvy generation of farmer.

For any VC who’s quick on the uptake, this confluence of micro and macro-level need on a wide-scale presents an opportunity to partner up with the right founders on unique and exciting journeys. While some interesting agritechs have popped up in different regions of the world – e.g. Kenya, Ghana, Egypt, Turkey, Pakistan, India, Korea, China, and USA – there is yet a lot of ground to cover in this space.

So how exactly can tech enable smallholders to thrive?

Markets

Bridging the gap between smallholder farms and end markets – where food retailers, processors, and consumers are buyers – is “ground zero” for empowering them.

Between them is a long chain of low-value-add middlemen, that buy and pool produce and engage in varying levels of wholesale commerce. In the absence of alternatives, farmers may depend on them for supply of inputs, credit, and harvest labour, as well as in crop selection. They are often powerful chain actors and command a high share of the margin stack.

Facilitating digital marketplaces for farmers and end market off-takers to transact directly is of considerable value to both chain actors. The value is in the ability to enter into forward contracts for the future delivery of produce – spot markets are only of marginal benefit in this case. Farmers enjoy more agency, visibility, security, and better pricing, while off-takers are able to tap into new supply nodes.

Digital marketplaces for agricultural inputs are also powerful enablers. Food safety requirements and contract specs place constraints on farmers in terms of inputs they can use – seeds, fertiliser, pesticide, herbicide, etc. Facilitating market access, product availability, guarantee of product authenticity, logistics, and price transparency, makes it easier for smallholders to do business in higher-value markets.

Fintech

There are 4 main financial products in which tech becomes meaningful for smallholder farms – credit, payments, insurance, and wallets.

Credit is a major concern for farmers given inherent working capital requirements and relatively long operating cycles, especially for smallholders who often have limited resources at their disposal.

The ability to accept digital payments is key to success, particularly when engaged in microcontract farming with large enterprises. This is a challenge for unbanked smallholders.

Ownership of a digital wallet/account allows unbanked farmers to receive and store inbound credit and revenue, as well as make digital payments for inputs.

Capital commitments, agricultural risk exposure, and limited disposable assets dictate smallholders’ critical need for adequate insurance cover, e.g. crop and livestock insurance.

Availing to smallholders access to such products – either as standalone fintech solutions or embedded within other marketplaces/platforms – bolsters their agency and ease of doing business in sophisticated end markets. The use of data to offer optimised credit pricing and insurance premiums are potential differentiators.

Precision agriculture

Low yields are characteristic of smallholder farms. A core contributor is their limited investment in technology – something that’s become mainstream in large-scale agribusiness.

Remote sensing via satellites, drones, or IoT devices is increasingly being used to monitor variables like soil health, crop health, temperature, humidity, crop growth, as well as to plan spatial crop structure. The data collected is powerful in enhancing farmers’ decision making, resource allocation, yields, and in turn their profitability.

Farm management SaaS allows smallholders to leverage that power in running their operation while avoiding hefty capex. This helps bridge the gap in their competitiveness vis-à-vis larger farms.

Supply chain

It is typical for agribusinesses to be vertically integrated. This gives them control over various nodes in the supply chain – e.g. post-harvest handling, storage, and transport – which is useful for minimising food loss as well as complying with food safety requirements of large and international markets.

The capex associated with that is beyond the means of smallholders – leaving them to deal with middlemen and/or informal service providers – which can cripple the competitiveness of their product and their bargaining power.

Traceability is also a key consideration in modern food systems – born out of increased food safety regulation and the heightened risk of hazard. This is basically the need to be able to trace a batch of produce back to its origin, and identify where it was grown, the inputs used, and the supply chain nodes that it’s passed through.

To comply, large agribusinesses either invest in traceability infrastructure or contract the services of expensive legacy providers – again to the exclusion of smallholders.

There are opportunities to commodify traceable and reliable post-harvest services (e.g. storage and transport) for smallholders, in the same vein as digital trucking, mobility, or fulfilment platforms. The blockchain is a resource for next-gen traceability.

This enables farmers to easily plug into higher-value markets, access previously out-of-reach ones, relieve themselves of post-harvest burden, and focus on maximising the success of their farm.

Social networks

Agricultural extension is the process through which farmers update their knowledge and practices in ways that help them be better and more successful at what they do. This can take place as farmer-to-farmer advice or through dialogue between farms and specialised organisations – often ones that help channel new information from the scientific community to the farming one.

The role of such networks is becoming more potent particularly in light of new knowledge about the interplay between agricultural systems and our natural environment, and the role of sound land, water, and soil management practices in preserving the integrity of both – practices that ultimately affect farmers’ livelihoods.

Digitising and formalising these networks/communities has the potential to maximise their reach, participation, and impact – think Stack Overflow for farmers.

Agritech platforms

Of the digital solutions mentioned, it is synergistic to bundle them together as part of multifaceted agritech platforms – or “super-apps”. There is precedent for various successful combinations. And their scope is not limited to only smallholders – there’s evidence to the potential of spill-over adoption amongst medium-sized farms as well.

Regional opportunity

So where is the opportunity most immediate?

There are 3 regions that stand out for having large numbers of smallholder farms contributing a sizeable share of food output and with ample potential for tech disruption. Those are Africa, Southern Asia, and South-Eastern Asia.

Egypt, Ethiopia, Ghana, Kenya, Nigeria, and Tanzania are standouts in Africa. While Bangladesh, India, Indonesia, Pakistan, Philippines, and Vietnam are ones in Asia.

Risk considerations

Agritech is a vertical where it makes sense to back founders with a deep understanding of farmers’ struggles – typically ones with domain expertise. The stakeholder landscape is complex, with a lot of moving parts, and it’s not easy for outsiders to crack.

Regional infrastructure, economics, and demographics also have bearing on product success. Internet coverage in rural areas, broadband speed, broadband cost, smartphone penetration amongst farmers, as well as their digital literacy pose challenges for adoption. The latter two relate to an ageing population – generational succession will take time.

Stakeholder collaboration and/or tech-based workarounds for these challenges – e.g. satellite broadband access, low cost smartphones with payment plans, low data intensity product design, user education programmes – are things to consider.

Final thoughts

Changing the status quo in this space is tough – more so than most other verticals. Yet the size of the problem and its potential impact make it an attractive one to take on.

Watch this space closely – it’s on track to become big in the next decade.

Saving is one facet of financial activity where fintechs in the Global South can create a lot of value.1 And it remains underpenetrated.

Households in the region have faced difficulty in building meaningful wealth over time. The real value of their savings is constantly being eaten into by high levels of inflation – disproportionately higher than those of their counterparts in the North. Weak currencies and import-dependent economies are often the root causes.

And there is practically little they can do about this phenomenon. Counteraction requires radical economic transformation and adept political leadership, both of which have by and large been elusive.

Meanwhile, their access to savings and investment products has been hindered by broadly 4 factors: rentier domestic financial institutions that have a vested interest in maintaining the status quo, a lack of a broad and diverse domestic asset universe, financial exclusion from big-ticket investments (e.g. real estate), and foreign exchange controls.

The capacity for tech to come in and transform the savings/investment industry is substantial. There’s been plenty of regional investment in availing credit products to consumers, and with good utility. But it needs to be balanced with universal access to savings products to be sustainable, particularly considering the rate of deterioration of wealth.

While the tech-driven democratisation of stock market investment in the North has been laudable on the whole, economies in the South are fundamentally different and availing wealth-related products to them requires a more nuanced approach. The quality of assets listed on domestic exchanges is something to consider, as well as their liquidity and their capacity to generate real returns over time – ones that exceed the rate of inflation. There is considerable utility in broadening the universe of quality investable assets domestically and availing access to foreign ones. Cooperation between fintechs and the regulatory sector on that is expected to be a core regional theme.

Democratising access to money- and fixed-income markets would also create meaningful value to savers. Treasury bills and bonds in the South commonly offer exceptionally high yields – ones that are more attractive than implied domestic stock market returns. This is particularly true in countries with chronic current account deficits and lots of government debt. These high yields are a bargain for domestic deposit-taking institutions (notably Egyptian banks) – they’re low risk, liquid, and the earned spread (after accounting for the meagre rate paid to depositors) is exceptionally lucrative. Availing households access to participate directly in such markets would create significant value for them.

The transient value of cash in these economies have led some to invest in alternative assets that are good hedges of inflation, such as land and real estate. Yet, given the low incomes and wealth disparities in the region as well as the ticket sizes in those markets, large swathes of the population are excluded from participation. A lot of value can be derived from giving these households access to upside in those markets. The blockchain and fractionalisation are technologies that allow this to be applied on a wide-scale.

Of course, in the long-term there is an onus on Southern governments to do better as well as on the international monetary system to level the currently-biased playing field (crypto is a compelling attempt) – such high inflation levels are not conducive to productive economic activity. Yet in the meantime regional households should not continue to be unjustifiably and disproportionately excluded from meaningful wealth building.

Keep an eye out for fintechs solving for these issues – they’re tackling a large and potent regional need.

Human capital has become an indispensable asset for countries today. The cross-border movement of goods and services that comes with globalisation pits the domestic output of economies constantly against that of foreign countries. To compete with the diversity and complex production of the most advanced economies, countries must maintain high levels of productivity and innovation. This goes hand in hand with having a skilled, knowledgeable, and technologically literate workforce. How well countries can maintain that dictates the level of income that residents can generate and, in turn, their standard of living.

Countries recognise this and invest in their people’s education. It is among the top priorities for governments worldwide. On average 14% of public spending goes toward education1, with the aim of making it universally accessible to citizens and attracting top talent to teaching professions. It is typical for countries to require by law that citizens attain education up to a certain level within formal education systems (up to age 18 in Egypt). This is largely to ensure their young population are equipped with the knowledge and skills necessary for them to be able to assert their ambition and realise their potential as productive members of society.

In spite of that, nations’ education outcomes vary in success. Egypt is a good example where they’ve fallen short of aspirations. The country ranks amongst the bottom third of 141 countries based on the skill level of their workforce and efficacy of their education systems.2 This deficit manifests itself in the economy in a multitude of ways, but most tellingly in the low levels of productivity and the low-level of complexity in the goods and services produced. Per capita GDP3, commonly used as a yardstick of productivity, is 26% lower than the world average and almost 5 times as small as that of the top 10 countries ranked by skill.45 In terms of the relative knowledge intensity of economies, and consequent complexity of their output, Egypt ranks amongst the bottom half of 127 countries surveyed.6

Education and teachers in Egypt

The reasons for this become clear when one considers the fact that the overwhelming majority of students in the country are excluded from access to top quality educators7. Research has shown that teacher quality is the strongest predictor of the quality of education.89 Public schools cannot afford to hire them and private schools, with their traditional cost structure, would have to charge high fees to be able to attract them – fees that are way beyond what is affordable for most households.

Similar to the ratio in countries with the most developed human capital, around 90% of students in Egypt attend public schools1011. However, public spending on schools per student in Egypt is one-thirteenth of what it is amongst the top 10 countries ranked by skill.121314 This is not by design – an overstretched state budget simply can’t afford more than that. Given that about one-half to two-thirds of this spending goes to fund teaching-staff compensation,15 this order of magnitude difference is an indicator of the disparity between the standard of living commanded by public school educators in Egypt and other countries with skilled workforces. Compared to other professions nationally, being a public school educator in Egypt places you somewhere within the bottom quarter on the income scale.1617 With such economics, public school careers are hardly well placed to attract the most ambitious talent.

Private schools are not generally much different. Using public data from one of the biggest private school operators in Egypt that caters to middle-income households,18 teacher remuneration is found to not be much more attractive on average than at public schools despite the higher school fees. On analysis, only around a third of the aggregate school fees paid are used to compensate teachers, with the lion’s share of the balance accruing to buildings and infrastructure-related costs, remuneration of non-teaching staff, taxes, and compensating investors for the capital they’ve put up. For the same schools, with the same quality of facilities and resources, to use competitive compensation to attract good teachers, it’s estimated they would have to raise their prices on average by at least a factor of 3. Such school fees would be out of reach for an overwhelming majority of the population.

Need for change

So how did Egypt reach this point, where a good education can’t be afforded by the state nor most households and the economy is in bad shape for it? This is more likely than not the result of a vicious cycle in which Egypt has been trapped for several generations. It takes only one generation of administrative neglect throughout which students receive substandard education to eventually bring down the relative competitiveness of an economy versus others, particularly those which have done better at training and skill building. In a globalised world, this likely results in less income accruing to domestic households in favour of foreign ones. In the absence of any counter measures, the lesser quality output of such an economy ultimately provides education inputs for the next generation – e.g. knowledge, culture, skills, curriculums, policy, pedagogy, etc. With a fast growing population like Egypt’s and rapid globalisation, such a problem snowballs with every turn of the cycle making it that much harder to break.

Why is this critical to address now? The economy is worsening and unless measures are taken to improve human capital the situation is unlikely to get any better. The adverse combination of low productivity and relative uncompetitiveness has put Egypt’s Balance of Payments (or BoP) in a state of ‘imbalance’. Despite sizeable inflows from tourism and Suez Canal receipts, Egypt ranks in the top third of 166 countries by the size of their BoP deficit19 as a share of GDP. Put simply, the country imports a large portion of goods and services and doesn’t generate enough income on aggregate to pay for them. Typically the shortfall is plugged by a combination of inbound remittances from nationals working abroad, foreign investment (recently volatile short-term debt), and grants.20 This precarious set up, coupled with the fact that the imbalance has been chronic, growing, and not met with sufficient gains in domestic productivity, is the reason behind the Egyptian pound losing more than three quarters of its value in the last two decades21. Recent developments place this issue in even sharper relief.

Another reason to address this sooner rather than later is the fact that the average household has limited freedom of choice when it comes to procuring education for their children – they are generally unable to maximise education quality per pound spent. They can choose between sending their kids to public schools at little to no cost but receive a subpar education, or spending a sizeable share of their limited income on private schools and receive in most cases only a marginally better product. This situation is made even graver when considering the unfavourable market dynamics at play. Demand for K12 education is understandably inelastic. This fact coupled with the fast pace of population growth is straining capacity on the supply side of the market. There is currently a shortage of at least 20,000 classrooms per year.222324 In the absence of a fiscally healthy state with a strong handle on education, a naturally opportunistic private sector is plugging the investment gap. Yet the combination of inelastic demand with chronic supply shortages (not to mention high barriers to entry) gives obscene bargaining power to school shareholders vis-à-vis households and educators. There is built-in upwards pressure on school fees as well as downward pressure on teacher’s salaries. Meanwhile there are few to no competitive pressures on private schools to deliver a best-in-class offering with the resources at hand – as long as they can deliver the minimum level of education required by domestic labour markets (no matter how poor by global and cultural standards) they will remain in business.

Food for policy

So what needs to happen in the immediate term to improve education outcomes at the national level with the given level of resources?

It’s important to conceptually decouple education from infrastructure. Egypt has one of the highest costs of capital globally. This means that for each unit of capital spent on school infrastructure, investors require more compensation for it than in most other countries. If anything this underlines the distinctiveness of economic levers in Egypt versus other countries, and hence the inefficacy of one-size-fits-all education policy – a more bespoke solution is needed. Building new schools needs capital for acquiring land, buildings, buses, desks, chairs, and equipment. We estimate that, of aggregate private school fees paid by households, at least 26% accrues to cost of capital (both equity and debt) and around 14% are used to cover infrastructure related expenses (e.g. depreciation, maintenance, and utilities). That’s at minimum 40% of expenditure by households that is potentially not being used to facilitate higher quality learning.

Schools are also ripe targets for disintermediation. They siphon a lot of value from national expenditure on education and what’s given back is not adequate. Take the teachers out of schools today, and their role is approximate to a combined registrar and day-care centre – they register students for exams and qualifications on behalf of third parties, and they watch over children and keep them occupied while parents hold their day jobs. These are necessary roles to fulfil, but can be done at a fraction of the cost of what schools effectively charge for.

The most important change that needs to happen is that teaching generally needs to be made more attractive as a career option for ambitious professionals. As custodians of the future prospect of the economy and its labour force, it is a critical occupation and it needs to be in better shape. Teachers ought to be empowered – they should enjoy more agency, more potential for economic mobility, and an expanded ability to reap the rewards of meritocracy. This is put in sharper focus when considering the current density of classrooms and chronic excess demand for education – there is a growing need for more teachers.

Technology and neo-schooling

How can these ideas be applied in the real world? With the use of technology and a new business model. A changed demography and economy, as well as the availability of new technology, has made obvious the fact that Egyptians’ need for education has evolved beyond the traditional school framework. Instead of dedicating expensive resources to schools, education can make use of underutilised existing ones. And instead of being school employees, teachers ought to be exchanging value with households directly, within the framework of an enabling ecosystem. It is easy to infer how this would yield better economics and improved education outcomes.

Consider the way stakeholders interact within the incumbent education system. Qualifications are generally awarded by the state, international non-profits, or companies.25 Qualification-awarding institutions (QAIs as they will be called here) typically also function as examination boards – they set exams, administer them, mark them, and distribute results. To sit for exams and be studying towards earning a qualification, students need to be registered with QAIs. Registration is typically done by schools on behalf of students. For schools to include a qualification amongst their offered programmes they first need to be accredited. Non-state accreditation is typically a process that involves a combination of applications, assessments, inspections, and the payment of fees. As part of a school’s accreditation process, training is typically provided to teachers employed by them – they are made familiar with the particular curriculum, pedagogy, and assessment methods. Fees are paid by schools towards non-state QAIs not just for the process of accreditation, but also for maintaining it, registering students, administrating exams, marking exams, sending out results, and the right to use branding. Fees generally recur either annually or by instance of service. The costs paid to QAIs naturally inflate the price of education at schools that offer non-state qualifications versus ones that don’t. For households, the education their children receive is a function of the school they’re enrolled in, the qualification they’re studying for, and most importantly the teachers hired by the school.

Replacing schools with digital platforms has the potential to streamline value exchange and maximise access to quality educators.

The business model of QAIs would need to change to serve beneficiary stakeholders directly. Accreditation of schools ought to become certification of teachers. And registration of students for exams and qualifications can be done digitally at the behest of households.

Digital marketplaces can match household demand for qualifications and education with supply of teachers. Many equilibriums can be found for pupil-teacher densities, teacher work hours, and course prices depending on the locale, demand for qualifications, household spending power, and teachers’ individual preferences. Scaling capacity to meet new demand would be as easy as new teachers onboarding to platforms.26 Households can also flexibly tailor aspects of their children’s education to their preference – if language ability is more important to them (as is the case in Egypt generally for English proficiency) they can spend more on higher quality language teachers without affecting the rest of their consumption mix. Quality control can be managed by the marketplace via ratings, reviews, and track records.

Non-curriculum education can also come into play here. Households that are willing and able can add off-programme subjects to their consumption mix as they see fit – subjects like financial literacy, hospitality ethics, sex education, and civics to name some examples. On the teachers’ side, this would expand market opportunities and create new jobs.

Under this model education would be real-estate-agnostic – it can take place anywhere. Households and teachers can agree on “classes” taking place in homes, outdoors, or dedicated learning facilities. Even if facilities were outsourced to third parties, better utilisation rates would improve cost economics. Scheduling would naturally be more flexible and is a component that would need to be managed well by platforms. Stay-at-home parents or gig economy workers can make incomes as chaperones for students while parents fulfil day jobs. At exam times, students can convene wherever QAIs administer them and sit for their tests. Results can be accessed directly by parents through digital platforms.

Evidence to the potential of communities to adopt a more decentralised “schooling” model, however limited, can be inferred from the rise of learning pods and micro-schools during the COVID-19 pandemic.

It is important to note that the fluency of technology would be key in attaining improved economics and learning outcomes. A platform business with low tech intensity would very likely devolve into yet another cost-bloated middleman. Only if platform software is efficient at automating most of what schools stood for would more value accrue to teachers from education spending. Traditional school-related expenditure for non-teaching staff, infrastructure, as well as costs of capital can be greatly reduced. Revisiting the example of middle-income private education, an estimated 45% of expenditure by households can be saved or spent on better teachers. If taxes on education were lifted27, the estimate would be more than 50%.

Stakeholder involvement

Three key stakeholders have a role to play in this scenario. Entrepreneurs need to create digital platforms that facilitate this new model of education, QAIs need to adapt, and the government needs to create a legal and regulatory environment that helps such platforms flourish. The government also needs to keep an eye on monopoly formation, as digital platforms naturally lend themselves to that. There must always be competition between at least two platforms to ensure the welfare of households and teachers is upheld. Given the critical nature of education to a nation, it would not be unreasonable for the government to attempt to create such a digital platform and offer it as a public utility. However, the execution track record of the public sector does not make this an encouraging idea. As an alternative to public education, the government can give subsidies to the most vulnerable households in the form of platform credits.

The government would also have a vital role to play beyond the scope of digital platforms (but in spite of them). They’d need to meet the increased adoption of decentralised schooling with commensurate investment in public facilities for sports, culture, and recreation. The logic is that students would still need outlets to express their physical and artistic selves in the absence of the school playground, art rooms, and theatres. The same goes for public libraries. This is particularly vital when considering the aggregate shortage of such public facilities in Egypt. It is worth noting that execution needs to happen with both universal access and good design in mind.

Final thoughts

It is important to consider the ideas expressed here in conjunction with other facts. The word education here refers to primary and secondary varieties (or K12). Their place as “ground zero” in a typical adult’s educational journey makes it necessary to examine them first in the context of poor education outcomes nationally. Tertiary (university-level) and technical education stages are worth at least equal scrutiny in that regard.

It is key to consider not just the question of who is teaching but also what is being taught. Curriculum content must maintain relevance and usefulness in the context of changing economics, science, technology, and culture. The dominance of software in business and hence the utility of software development skills is a good example of such change.

Needless to say, human capital is only part of the story. An educated workforce can’t be effectively mobilised without the presence of an enabling environment. This includes the rule of law, an impartial system of resource allocation, and strong public policy. Such an environment is necessary to also avoid human capital flight.28

The intention here is not to provide an ideal imagining of schools nor to reinvent educational paradigms. But rather redraw the system through which education value flows in such a way that would improve the Egyptian status quo (and that of any countries with similar economics). There have been efforts by the state to do so but not radical enough. The same can be said of the current cohort of domestic tech start-ups – they are trying to create value within the broader education market,29 yet schools remain largely undisrupted. It is critical for change to happen to preserve and promote the welfare of the domestic economy, particularly given the low skill levels and the chronic net flow of value out of the country. The ideas presented here are done so in the hope that they’ll help in bringing that change about.