Could AI Empower the Bottom of the Pyramid?

In a world where AI radically reduces the cost of knowledge work, what does this mean for people on low incomes?

Developing services for consumers on low incomes

Many years ago when I was a management consultant, I worked with a very large health insurer on new product development. Our focus was how a health insurer could use its assets and scale to open new markets. My interest in social impact meant asking what products a private health insurer could develop for people on low incomes. We identified that many people on low incomes in Australia struggled to pay for major dental work because dentistry is not covered by the universal public health system (Medicare).

Having poor teeth makes it harder for somebody to go get a job. There's evidence (1) that dental aesthetics (how your teeth look) lead people to draw conclusions about the social characteristics of a person. If people tend to judge you on your appearance then having visibly poor teeth can have a real impact on self-confidence and then that becomes an additional barrier to building trust and therefore getting a job. So enabling people on lower incomes to fix their teeth can have positive social impact.

In that project we developed an innovative new dental savings product that was intended to help people on lower incomes afford dental care. A sort of savings plan combined with the power of the private health insurance to negotiate lower costs could mean that we could significantly reduce the costs and risks around much needed dental work.

But there's a 'but'. It may not surprise you that the project was cancelled. While the health insurer had a health-based mission, this was still a commercial business. Executives with limited time naturally preferred to focus on creating and selling products that had higher profit margins, and therefore innovation effort was focused more on the wealthy market where margins for insurance products were much higher.

The 'bottom of the pyramid'

The book 'Fortune at the Bottom of the Pyramid' (2) was written in 2004 and argued that businesses could develop profitable products for the large group of people on low incomes around the world. One of the famous case studies used was of Unilever, the giant Anglo-Dutch consumer products company that experimented with packaging its shampoos into smaller packets to enable people on low incomes in India to buy small amounts as and when they could afford it. Yet it's still common for the needs of people on lower incomes to be overlooked and 'under-served' by business. It turns out to be very hard to reduce costs of many products and services to the level people on low incomes can afford.

AI as a disruptive innovation system for social problems

These stories are helpful frameworks for thinking about how artificial intelligence (3) could create opportunities to develop new products and services for people on lower incomes (with a focus on developed economies).

We know from history that any technology that increases productivity significantly will lead to much lower output costs. And when the costs of a product drop significantly, it often means the market expands. The combination of steam power and the invention of the 'power loom' in Manchester in early 1800 led to an estimated 20x increase in the productivity of a weaver's labour, resulting in the real price of cotton cloth falling by approximately 85% to 90% between 1780 and 1850, (4). But this also led to a huge impact on labour (less weavers needed) and social upheaval as people left the country and moved to cities.

My prediction is that social entrepreneurs will leverage AI to reduce prices of many services (perhaps by similar percentages as the cost of cloth fell during the industrial revolution) and therefore rapidly expand access to services that were previously unavailable to people on lower incomes. At the same time, the huge productivity gains will impact employment in knowledge industries. It's complex because both positive (cheaper services for people on low incomes) and negative (reduced labour demand for knowledge workers) will happen at the same time. This article focuses on the positives for the 'bottom of the pyramid'. Here are just two areas of likely impact:

A) Financial services. Think about that dental health savings product that I described. If AI gets to a point where it can handle much of the legal drafting, compliance work, actuarial modelling, and customer service at very low cost, then the opportunity will exist for an entrepreneur to build such an insurance product or solution without needing a big corporate structure behind them. The AI could potentially do the legal and compliance work required to get a product registered with the relevant authorities. It may be able to develop the marketing and build connections to the right buyers, as well as run the customer service and actuarial modelling required to price and develop the product. Imagine a dental insurance product that cuts costs by 60% or 80% - that kind of innovation could make such a product affordable to people on lower incomes for the first time. Or imagine an AI mortgage broker who negotiates thousands of dollars in mortgage savings. Or an agent that applies for the right government benefits for a person with a disability. Or negotiates lower rates for car insurance.

B) Legal services. In the old days if a person on a low income got into legal trouble, they would often have to rely on a publicly funded lawyer who may not have the time to prepare for their case properly. In a world where AI radically reduces the cost of legal analysis and advice, a person on low income may be able to use such a service to develop a better case at very low cost (almost free?) and therefore improve their chances of winning in court - say getting unfair debts wiped off or reducing a criminal sentence. What happens when anyone can afford a 'good enough' lawyer? A community of people who are victims of a corporate environmental crime could band together and develop or even run their own case using AI. Perhaps we'll see an explosion in class action lawsuits and hundreds of "Erin Brockovich" stories will be written around the world.

Of course, many of these knowledge services are regulated for good reason, and regulatory frameworks will need to evolve to enable AI-empowered alternatives while still protecting consumers. A human that oversees these AI agents is still likely to be needed to get approvals and take legal responsibility - just as the power loom still needed an operator, but the machine and the human together did the work of 20 individual weavers.

There's no doubt AI will have a significant effect on jobs for people doing knowledge work. That deserves serious focus and likely regulation and intervention by governments to manage its effects. A lot is written about this risk, but we sometimes may overlook how AI has the potential to create real value for people on low incomes. What’s important is we build a generation of entrepreneurs who have a mission to add value to the lives of people on low incomes and develop clever and creative solutions using AI to expand access to services, but do it in a way that delivers real value to those people and avoids exploitation. 

(1) Dentofacial Aesthetics and Quality of Life, Ulrich Klages & Andrej Zentner 2007

(2) The Fortune at the Bottom of the Pyramid by C.K. Prahalad, 2004 - the thesis is that firms can profitably serve very low income consumers (e.g. a case study on Unilever in India), but there are also criticism of this theory as successful development relies on other factors such as good governance to work, not just private enterprise.

(3) When we talk of artificial intelligence in this context we mean Large Language Models and 'agents' that can do work on behalf of a user (such as use software, research information, or do specific tasks such as review a contract or develop a financial model). As the LLM models and tools keep advancing, we can expect them to get better at doing many knowledge-intensive tasks.

(4) The Hand-Loom Weaver and the Power Loom: A Schumpeterian Perspective, Robert Allen, 2017