Page 284 - The Digital Financial Services (DFS) Ecosystem

ITU-T Focus Group Digital Financial Services
Ecosystem

Figure 22 – Chat providers disruption to the mobile ecosystem

Chat platforms could also reduce the role of MNOs. MNOs typically charge separately for voice, SMS, and data
services (visiting social networks, sending email, etc.). Chat platforms also provide voice and messaging services
to their users on top of the MNO’s data service. This arrangement reduces the user’s costs to the detriment
of the MNO. MNOs may become commodity data pipes if this trend continues.

I.4.4 Transformational benefits of artificial intelligence (AI)?

Today, chatbots often operate within a structured framework. Users make requests via a menu ("Shop for
shoes," "Shop for shirts," etc.) and chatbots retrieve ‘canned’ responses from a database. This approach
provides a predictable customer experience but does not scale well – use cases are limited since businesses
must determine scenarios and appropriate responses.

Sci-fi movies depict a utopian future in which anyone can ask any question and receive a useful reply. Users
have flexibility in how and what they ask, with voice recognition removing literacy requirements. Responses
are not limited to pre-defined answers. For example, a consumer might ask "Where can I find a replacement
part for my water pump?" After some back and forth clarification, a computer might return "The replacement
handle is available immediately at Paul’s Irrigation Supply in Nairobi for 200 Ksh, but if you can wait until next
Wednesday, it will be available at the local K&M store for only 180 Ksh."
Two technologies provide a step toward this AI vision: natural language processing (NLP) and machine learning.
NLP goals include natural language understanding (deriving meaning from human or natural language input)
and natural language generation (communicating back to the user). This extremely complex field includes
topics such as speech recognition, name entity recognition (identifying words that are people or places), and
part-of-speech tagging (identifying whether words are nouns, verbs, etc.). In the early days, most NLP systems
were based on complex sets of hand-written rules (i.e., if you detect the phrases "I need help" or "customer
service" or "live person" route the caller to a customer service agent.). Machine learning has made the rule
development more robust – e.g., analysing historical data to find phrases that predict the need for a customer
service agent.

Despite the progress and level of effort, fulfilling the AI vision is a long way off. Imagine the user frustration if
a Google search returned only one sentence from one article. This scenario represents how far we are from a
true AI vision. Fortunately, the AI promise is not an ‘all or nothing’ proposition. The journey will likely involve
incremental change, producing incremental benefits.

I.4.5 Consumer and merchant value proposition?

Chat platforms are strategically important to social networks, but do they help consumers and merchants? The
AI vision has obvious appeal because it lets people communicate in a familiar and natural manner. But, utopia
is a long way off. In the meantime, is the current version of chat platforms providing value?

Facebook’s CEO Mark Zuckerberg argues that chat offers a superior user experience, "No one wants to have
to install a new app for every business or service that they want to interact with. We think that you should

256