Derin™ Immunization & Family Planning in Karachi

Conversational AI for the hardest-to-reach research populations

Product: Derin™ 

Summary

Surgo Health’s AI behavioral intelligence platform Derin ran multilingual voice and text interviews with mothers of young children in Karachi, capturing firsthand accounts of how families make decisions about childhood immunization and family planning. Conversations revealed the behavioral drivers behind health decisions, including vaccine misinformation, spousal and household barriers, mistrust of health workers, and mobility restrictions. Derin reached women that traditional methods can struggle to reach, while surfacing both quantitative and qualitative depth.

What Traditional Measures Miss, and Why It Matters for Health Organizations

Traditional research approaches often struggle to balance scale, speed, and depth– especially when engaging populations with varied literacy levels, limited digital access, or sensitivity around disclosure. Derin bridges this gap by enabling structured yet conversational engagement that captures both quantitative signals and qualitative context in a single flow. By combining survey-grade structure with interview-level depth, Derin helps organizations:

• Reach diverse populations, including those with limited device or digital access
• Elicit candid, high-quality responses on sensitive topics through privacy-aware design
• Generate rich behavioral data that reflects both what people do and why they do it
• Accelerate formative research timelines from months to weeks without sacrificing rigor
• Scale human-centered insight generation across geographies and program types

This behavioral intelligence helps health organizations move beyond traditional tradeoffs between scale and depth, and instead build faster, more inclusive, and more behaviorally grounded evidence for decision-making.

Challenge 

In low- and middle-income settings, formative health research often relies on focus groups and enumerator-led surveys that are slow, costly, and hard to scale. To design effective interventions and engagement strategies, program leaders need to understand how information gaps, social norms, trust, and access barriers shape people’s behavior. Reaching women with low literacy, no personal device, or restricted mobility, while gathering actionable insights, remains a challenge in the field.

Approach

Derin was used to hear in-depth perspectives from women in Karashi that traditional survey methods can struggle to reach. It ran in Urdu by voice and text, leveraging a friendly AI persona, onboarding video, and privacy gates before sensitive questions. A tiered support model paired self-service for women with smartphone access and digital literacy alongside enumerator support for those without, extending reach across segments. Derin’s topic modeling analysis surfaced key themes and drivers across responses.

Results 

1. Conversational Depth

Derin conversations produce rich, in-depth findings. Each conversation covered between 8 and 11 behavioral themes, spanning both outcomes and the relevant drivers behind them. The family planning section alone yielded the equivalent of 30-60 quantitative variables per interview. Open-ended text answers produced cleaner data than long multiple-choice lists, which confused participants with reading difficulties. Women also responded to the multiple modalities offered by Derin, using voice responses to explain their experiences in greater detail.

8-11 Behavioral themes surfaced per interview

30-60 Variables covered in each family-planning conversation

21 Words per voice response, 4x longer than text

2. Adaptive to Varied Needs 

Scaling survey responses requires matching support to the literary, device, and access profile of each woman. No single delivery model reaches everyone, so participants were sorted into four segments, with trust cues held constant and hands-on support provided when needed. The most enabled segments completed digitally on their own or with light support. The rest required assistance during recruitment, guided onboarding, or full enumerator-led delivery with a provided device.

35% Highly enabled: digital self-service

32% Partially enabled: digital with light-touch phone support

16% Support enabled: partial assistance during recruitment

17% Support and access enabled: enumerator-led, device provided

Derin reveals that beneficiaries prioritize Medicaid eligibility over certain job opportunities. Jobs offering health insurance or paying enough to purchase a marketplace plan are thin, and respondents feel disincentivized to accept new roles or work more hours. For those who turned down job opportunities to maintain eligibility, losing coverage now will mean rebuilding both their healthcare access and their career path.

3. Earned Trust

Derin prompted candid disclosure on topics typically considered taboo. Trust-building features were built into Derin proactively. The AI interviewer adopted the persona of  “Mobile Maryam,” a Pakistani woman designed to feel like a trusted friend, featuring a matching WhatsApp profile photo. Onboarding video and prompts were tuned to this persona. Before the family-planning block, explicit privacy gates advised women to lower their volume or move somewhere private. Participants cited the privacy nudge as a reason they felt in control when answering sensitive questions.

92% showed active engagement

84% completed the full interview

79% consented to share immunization status and intent

See How Derin Turns Conversations Into Action

Move beyond what happened to understand why it happened. See how Derin uncovers the behaviors, motivations, and barriers that traditional analyses miss.

Talk to an expert and see Derin in action.