Researcher and SMB using the Dukawalla platform

Dukawalla

A voice first AI data assistant that enables small & medium businesses (SMBs) to collate their business information, and glean rich insights

Background

Based on a research study on Digital Transformation among small and medium businesses (SMBs) in Africa, one of the biggest needs of SMBs is rich insights that can equip them to make better decisions easily. SMBs in Africa are currently using fragmented data collected from different physical and digital channels to make decisions. However, many SMBs are still struggling to make business decisions that meet their customers’ needs and enhance their businesses’ productivity. To glean rich insights, SMBs must first collect and manage their data in one place. However, collecting and managing data using the digital tools currently available is difficult for SMBs; many find that the products available have a steep learning curve, complex interfaces which they do not have time to understand and often do not fit into their ways of working nor their technological environments.

Business owner recording product for sale via voice interaction on Dukawalla

The big idea

To cater to these and similar research insights, we built the research prototype Dukawalla. Through this prototype we explored and tried to understand how a data management solution with voice as its primary interaction point could enable SMBs to leverage their data to increase productivity and grow their businesses.

Dukawalla is a mobile-first research prototype that enables data capture via speech models. Through voice interaction it leverages Large language Models (LLMs) to transform unstructured data to structured CSV. Dukawalla is also powered by LLMs and Image Generation Models (IGMs) to present bite sized insights to the SMB to allow data driven decision making. The prototype relies on Excel as a Backend service for data collation and creates a user experience based on socio-tecture.

Leveraging Large Language Models for small businesses

flow chart showing the different technologies integrated into dukawalla - Voice to record and ask questions about your data​, Language understanding. Structure conversational data into CSV​, Collate business data in Excel through Microsoft Graph APIs ​, Lida to fenerates data summaries and recommended visualizations with explanations on questions asked​ all creating the easily comprehensible and real-time summaries of business performance​ in Dukawalla

Prototype testing

In 2023, we engaged with small and medium-sized businesses in Nairobi to test our research prototype. This 2 month engagement yielded valuable insights and identified some areas for improvement. The businesses appreciated the ability to quickly generate graphs and infographics, aiding in swift decision-making. Dukawalla provided crucial support in data management, a common challenge for SMBs. However, we encountered difficulties with the speech model, which affected the accuracy and effectiveness of natural language processing using SMB data. Moreover, voice interfaces were not consistently suitable for SMBs due to environmental factors, privacy concerns, usage timing, and social norms.

Researcher and small business user using Dukawalla app

By designing integrated technologies with and for SMBs in Africa, we are more likely to design more fitting communication and collaboration technologies that will empower millions to make data driven business decisions.