Listening to customers at scale

As rural tap coverage in Assam crossed 80 percent under the Jal Jeevan Mission, attention started shifting from building infrastructure to ensuring reliable everyday service. Supply-side systems could track assets and scheme status but there was limited visibility into how households experienced water supply. Issues related to irregular supply, inadequate quantity and concerns about water quality often remained anecdotal or reached officials too late. Existing grievance mechanisms were important but largely reactive and fragmented. Arghyam worked with the Government of Assam to test a structured way to listen to citizens at scale. CSAT was designed to generate demand-side evidence that could complement administrative data and help officials identify what required attention, where to prioritise and how to track improvements over time.

CSAT was designed for accessibility in rural settings where digital literacy varies and smartphone-based surveys can exclude users. Voice calls were chosen because they are familiar and reduce the effort required to participate. A short survey was developed covering five aspects of service delivery: regularity of supply, consistency of timing, adequacy of quantity, perceived water quality and overall satisfaction.

The pilot tested two modalities, IVR and conversational AI, to understand response rates, completion rates and the quality of data generated. Language localisation and speech design were prioritised to make the experience natural and reduce drop-offs. People Plus AI supported the design and testing of the AI voice approach and Reverie supported language and speech localisation.

During the pilot, 18,000 calls were placed. The IVR channel recorded a 61 percent response rate, with 10 percent of respondents completing the full survey. Data was analysed to identify patterns that could support action, including how regularity of supply influences overall satisfaction. Findings were structured in a way that state and district teams could interpret quickly, rather than as a one-off research output.

In parallel, the work explored what it takes to institutionalise citizen feedback as part of routine governance. This included considering how feedback should flow to the right level for follow-up, how officials could use citizen data alongside scheme data and what it means to close the loop with communities. The pilot demonstrated that citizen voice can be captured at scale and that with the right design it can become a practical input for improving day-to-day service delivery.