The most expensive part of any global health study is often the human conversation. You need to find people, train interviewers, translate questions, and then spend weeks or months in the field, all before a single insight is logged. It is a process measured in person-months and carbon, and it struggles to reach the most remote communities. HumanTruths, a Cambridge startup founded last year, is betting that an AI voice agent can do it faster, cheaper, and in more languages.
Its first large-scale test is underway: a study of more than 2,000 Kenyan citizens, focused on understanding childhood immunization hesitancy. The work is backed by a $684,000 grant from the Bill & Melinda Gates Foundation, and the platform promises to turn those thousands of conversations into structured behavioral insights in days, not months [Harvard Innovation Labs, December 2025] [vcbacked.co, October 2025]. For organizations that have spent decades and billions on surveys, it is an attempt to swap jet fuel for compute cycles.
The wedge: Replacing the field researcher
HumanTruths is not building another dashboard for quantitative data. Its wedge is qualitative, emotional, and linguistic. The platform uses AI voice agents to conduct interviews via phone, SMS, WhatsApp, and other channels, supporting over 50 languages [humantruths.ai]. It then applies behavioral science frameworks to parse the conversations, looking for the unstated beliefs and social norms that drive decisions around health, finance, or governance.
The bet is that this method unlocks populations and speed that traditional research cannot match. Sending a team to a rural village is costly and slow. An AI agent can call 100 people in that village in an afternoon, ask follow-up questions in their local dialect, and never get tired. For partners like the UN, WHO, or Gavi, the Vaccine Alliance, the appeal is clear: higher fidelity data from harder-to-reach communities, at a lower operational cost [LinkedIn].
A team built for the intersection
The founders assembled a team that reads like a Venn diagram of the problem space. CEO Ceren Koca was previously a Principal Strategy & Insights Lead at Google in the Asia-Pacific region, and has advisory experience with UNFPA [Stealth Startup Spy, August 2025]. Her co-founder, Luigi De Angelis, is a physician-epidemiologist and a co-founder of the Italian Society for AI in Medicine, currently a public health resident at the University of Pisa [Startup.Stream].
This combination of commercial behavioral science, global development networks, and clinical public health expertise is the company's foundational asset. They met and won an AI Venture Studio Demo Day before formally starting the company in 2025 [LinkedIn, January 2026]. The early grant from the Gates Foundation suggests their thesis resonated with a funder who has seen every iteration of global health research.
| Founder | Role | Key Background |
|---|---|---|
| Ceren Koca | Co-Founder & CEO | Ex-Google Principal Strategy & Insights Lead; UNFPA advisor; Harvard Innovation Labs affiliate. |
| Luigi De Angelis | Co-Founder & Chief Research Officer | Physician-epidemiologist; co-founder, Italian Society for AI in Medicine; public health resident. |
The grant-funded proving ground
With 1-10 employees, HumanTruths is in the build-and-prove phase [LinkedIn]. The Gates Foundation grant is its sole disclosed funding, a non-dilutive runway that aligns with its social-impact mission. The capital is being used to staff up engineering roles for backend and cloud infrastructure, indicating a focus on scaling the core platform [LinkedIn, February 2026].
The Kenyan immunization study is the most concrete proof point in the public record. If successful, it serves as a blueprint: a large-N, AI-conducted qualitative study that produces actionable insights for a major global health initiative. Success here would be a powerful case study to take to other NGOs, development agencies, and mission-driven corporations.
The risks of a novel methodology
The model is elegant in theory, but it introduces new questions that traditional research does not face. The company's ability to scale hinges on three core challenges.
- Cultural and linguistic nuance. An AI model trained on one dataset may misinterpret local idioms or social cues. The platform's claim of 50+ language support is a technical feat, but true understanding requires deep cultural context.
- Participant trust. Will people speak as openly to a machine as they would to a human interviewer from their community? Building initial trust and ensuring ethical consent in an AI-mediated conversation is an unsolved behavioral challenge.
- Commercial transition. Grant funding validates the mission, but long-term sustainability requires paying customers. The pivot from prestigious NGO partners to enterprise contracts with recurring revenue is a classic hurdle for social-impact tech.
The founders' backgrounds suggest they are aware of these layers. De Angelis's medical ethics training and Koca's work in cross-cultural behavioral insights are direct responses to the first two risks. The third, the business model, remains the open question.
What to watch in the next 12 months
The immediate milestones are clear: execute the Kenyan study, publish the findings, and convert that credibility into initial paid contracts. The Gates Foundation grant provides an 18-24 month runway (estimated), which means the clock is ticking to demonstrate a path beyond philanthropic support.
Key signals to watch will be any shift in hiring toward sales or customer success roles, the announcement of a priced equity round with venture investors, and the details of any follow-on studies. A partnership with a government health ministry or a large philanthropic contractor like FHI 360 would signal that the tool is moving from pilot to program.
On a unit economics level, the calculation is straightforward. If a traditional field study involving 2,000 subjects costs $300,000 and takes four months, and HumanTruths can deliver comparable insights for $100,000 in two weeks, the value is undeniable. The savings are not just in dollars, but in the carbon footprint of eliminated travel and the accelerated timeline for intervention.
The incumbent HumanTruths must beat is not a software company. It is the entrenched, person-intensive methodology of international development research itself,a globe-spanning network of consultants, translators, and enumerators. The bet is that in the race for understanding, algorithms can be more humane, not less.