Bandhu leverages conversational AI and Glific to connect migrants to affordable rental housing


AUGUST 23, 2023


This blog is written by Jacob Kohn, Bandhu

Bandhu is a technology platform that connects blue-collar workers across India to jobs and affordable rental housing opportunities. They have three smartphone apps and adopted Glific’s platform for WhatsApp two years ago, being one of the first partners to do so. 

Bandhu uses Glific to help blue-collar workers access their tech platform and get registered, find a job, or find housing. Although it has registered over 95,000 users on its platform to date, Bandhu found that a key obstacle to rapid scaling was the learning curve required for a worker to use a new smartphone app. Workers would need to be comfortable downloading the app from the Play Store, signing up for an account, and learning search and transaction behaviours within a new app. 

By contrast, many workers are already familiar with apps like YouTube and WhatsApp; Bandhu found in its field research that some workers even term any smartphone a “WhatsApp phone.” This finding pointed to a key opportunity to accelerate the adoption of their platform through WhatsApp itself. Further, WhatsApp was an ideal platform to create a digital “path” to guide migrants to urban housing, documentation and financial inclusion–migrants who often struggle even to prove their existence.

In 2021, Bandhu released its first iteration of a WhatsApp chatbot interface, powered by Glific. This effort drove a 20% increase in app downloads and registrations, and enabled workers to get connected directly to a member of Bandhu’s support team. However, the solution ran into scalability challenges: workers were limited to particular conversation flows, and these predetermined pathways were cumbersome to maintain and translate into all the possible languages of users throughout India. Additionally, these workers, being newly digitized, did not yet fully trust making transactions digitally. 

Bandhu required a scalable, flexible solution that allowed workers to access Bandhu’s services in their native languages, in a conversational format that would also enhance users’ trust in technology and drive digital-first behaviours. 

Bandhu + Glific + Jugalbandi

A compelling opportunity to develop this solution arrived in April 2023 when Bandhu became a beta tester for Jugalbandi, an AI-driven retrieval engine that uses LLMs specialized for Indian languages. Around this time, Glific was also beginning to realize the potential of Jugalbandi for a variety of use cases for its partner NGOs.

What made Bandhu’s use case unique was its ability to derive a knowledge base from not just from a static datasource, but Bandhu’s live database of housing stock itself. 

Bandhu’s integration with Glific and Jugalbandi solves four digital trust and accessibility-related pain points for a worker: 

  1. Entry point: Ease of accessibility (i.e. most workers have WhatsApp already installed as a smartphone app and trust it as a communications platform)
  2. Conversational nature: Ability to understand a user’s need whether their query is structured or unstructured, in any of India’s 12 major languages
  3. Live results: Ability to query Bandhu’s live housing database and fetch logical options based on the worker’s query
  4. Literacy-agnostic: Ability to parse both text and voice notes, and return a response in both formats

What Bandhu has built

A migrant can interact with Bandhu’s AI chatbot by sending a message to the chatbot (+91 9104000898) and entering a special AI-enabled flow. The user can then send a text or voice note stating their requirement, i.e. “I want to find an apartment in Ahmedabad with a budget of less than Rs 10000 per month.” This request is then parsed and passed to Bandhu’s housing database, generating a response with various housing options and their details. 

To minimize decision fatigue on the part of the user and drive quick decision-making, Bandhu returns a WhatsApp message featuring several results at a time in the form of a special link to their website. Each result contains details of a particular housing offering that satisfies the user’s requirement. From there, the user can submit an inquiry and receive a callback from the Bandhu team to learn more about the offering, arrange a property visit, and move forward with the transaction. 

Measuring impact at scale

Bandhu’s solution also actively generates data that can be used to further hone the chatbot’s responses, using Glific’s Google Sheets integration to track user responses and rate the accuracy of these responses. Bandhu tracks the number of users engaging with the chatbot, using it to find housing, and contacting Bandhu to close the deal. 

Looking ahead

While this solution is rapidly gaining traction on-ground in Ahmedabad and other parts of western India, its ability to parse queries and respond in south Indian languages such as Tamil and Malayalam are currently less accurate due to the dependency on LLMs and limited training data of these models for such languages. However, this should improve over time as more data is added to Jugalbandi’s integration. 

Bandhu is scaling its solution to 12 major cities in India, and this offering will accelerate their ability to reach users in any language, anywhere in India. Once users are shown to be using Bandhu’s chatbot to find and secure housing, the next steps will include making secure transactions via WhatsApp and offering additional opportunities that reward the user for digital-first behaviours—such as loan offerings. 

Bandhu hopes to add further integrations with social media to supplement its growing database of affordable rental housing offerings. 

Leave a Reply

%d bloggers like this: