June 15

Oliver Gilbert

Getting 4 Billion New Internet Users Online: Deploying Conversational AI at the BoP

Do you remember JARVIS, the computerized personal assistant who helps Tony Stark manage his complex super hero affairs in the Iron Man movies? With the rise of conversational Artificial Intelligence (AI) like Google Home or Amazon Echo, or the smartphone-based Viv, Siri, Cortana, etc., similar JARVIS-like services may be available in the not-so-distant future.

Yet to date, the impact of conversational technology – and popular discussion about its use – often remain limited to modern conveniences like ordering household supplies or receiving restaurant recommendations, reminders and hands-free directions.

While it’s helpful for everyday tasks, the technology that underlies these services can be put to more ambitious uses. Indeed, it can amplify existing impacts and overcome significant technical barriers to unlocking commercial, educational and connectivity services for the next four billion emerging mobile users in developing markets – people who will soon access the internet for the first time.


Mobile Penetration vs. Mobile Literacy

Global smartphone penetration is skyrocketing, and the 2.2 billion people using mobile internet in 2013 are expected to increase to 3.8 billion by 2020.

Nonetheless, there is a significant difference between mobile penetration and mobile literacy.

Billions of people in developing countries lack basic mobile internet literacy skills, defined by the GSMA as the ability to “set-up, search and navigate, consume and evaluate, create and share, and configure and manage.”

How big is this difference? A Grameen Foundation study found that only 36 percent of women in India know how to independently use a mobile phone. Similarly, among respondents interviewed by the GSMA, 76 percent of women and 61 percent of men in India reported that they need help using mobile internet, and that “many admitted they were not completely sure what the internet was, or what it could be used for.”


Barriers to Digital Literacy

The three top barriers to adoption – beyond device cost and available coverage – are technical literacy, difficulty interacting with English text, and fear of making mistakes and losing money.

The lack of technical literacy strands users on what GSMA describes as “application islands,” in which users understand how to use a particular app, like Facebook, but fail to comprehend how processes, symbols and capabilities can be applied to other applications. This results in users genuinely believing the application is the entire internet.

Furthermore, English text in phone interfaces, online content and search results limits users’ capacity to toggle device settings or input and understand search queries. And beyond English barriers, many non-Roman character languages – like Hindi, Myanmar and Bengali – are difficult to input on standard keyboards for new and existing users alike.

Lastly, new users at the base of the pyramid (BoP), who often earn less than US$5-7 per day, are reluctant to experiment with phones or the internet because missteps waste data and limited income.

These under-addressed barriers limit the expansion of mobile internet adoption, access to information, and needed social services in developing and emerging markets.


Voice-First User Interfaces for the BoP

Utilizing conversational AI as new users’ primary online interface could enable new BoP users to leapfrog the technical challenges that are currently limiting mobile literacy and adoption.

Prioritizing conversational user interfaces on new smartphones would lower barriers to access by:

  • Enabling local language voice transcription, search and review, including for development-focused services like mobile maternal health care information for new mothers, or agricultural best practices for farmers
  • Troubleshooting steps to open personal or financial accounts for unbanked rural consumers
  • Facilitating interactive education opportunities for students, similar to the 31 percent of students in developed markets who already use voice assistants for homework help

Leveraging voice isn’t new. Interactive voice response systems (IVR), like automated customer hotlines, have been utilized for years and have even been applied to delivering entertainment and education to the BoP. However, IVR campaigns can be technically complex and expensive for NGOs, small businesses or startups focused on the BoP to use at scale.


Obstacles to Implementation – And Solutions

Existing voice assistants currently maintain the tech capacity and language resources to support over 40 major languages – but even with dozens of languages available, many at the BoP could be excluded. And companies frequently lack business models to justify prioritizing new investments in under-represented languages and subnational dialects. What can be done to address this gap?

Similar to government investments in roads or schools to promote shared public services like commerce and education, governments and donors can support technical training programs in emerging market developer communities to build upon open APIs. They can provide grant funding to develop underrepresented language packs as a public investment to accelerate mobile usage, education and commerce at the BoP.

As for costs, the company responsible for building Siri reportedly raised US $24 million of venture capital in 2010 prior to being sold to Apple for US $200 million. To put these numbers into perspective, in 2014 the world spent US$135 billion on development aid. So even if new tools were designed from the ground up, the expense of creating a voice assistant couldn’t be described as unreasonable in comparison. And in practice, the cost of building new language support for existing platforms would be far less than building a platform from scratch, with expenses on par with those of other multi-year development programs.

Best of all, once these voice assistants are developed, they’ll remain as accessible assets that lower barriers to entry for other non-profit and private sector organizations building leapfrog services for the BoP. Whether it’s called JARVIS, Siri or Viv, wouldn’t it be cool if this technology actually helped increase equality of global mobile access and information, in addition to just ordering take out?


Oliver Gilbert directs Next Billion, a Singapore-based social enterprise dedicated to engaging the next billion emerging mobile consumers.

Note: The Singapore social enterprise is not affiliated with



Top image: The Amazon Echo voice command device. Photo credit: Pierre Lecourt via Flickr