Healthcare providers are often involved in helping clients consider, choose, and adopt contraceptive methods.
This includes new methods like DMPA-SC, the pre-packaged, single-use contraceptive that can be self-injected and is effective for three months, thereby reducing the return visits a user needs to make to the clinic.
Satisfaction around self-injected DMPA-SC is very high once initiated, but many users have to first overcome their fear of needles or a lack of confidence in their ability to inject it to start.
Advances in AI present an opportunity to develop highly interactive, personalized training tools that can help providers strengthen their counseling skills for more nuanced scenarios. For example, an AI-powered chatbot performed well supporting healthcare providers in Benin.
AI-Powered Coaching for Healthcare Providers
Jhpiego and Dimagi designed and tested an LLM-powered chatbot as a proof-of-concept to strengthen providers’ capacity to help clients navigate needle fear and confidence-related issues with self-injection.
The interactive chatbot was built on Dimagi’s experiences to help providers navigate clients’ vaccine hesitancies. This chatbot specifically targets healthcare providers who counsel clients on DMPA-SC.
The DMPA-SC Coach chatbot was built using Dimagi’s Open Chat Studio and developed through a collaborative process using DMPA-SC self-injection training materials from around the world – including supplemental content for addressing confidence and needle fear – and by involving family planning healthcare providers from Cotonou, Benin.
How the DMPA-SC Coach Engages Users
The DMPA-SC Coach was made available for testing via browser and WhatsApp, and designed with key features including:
- Interactive Training Exercises: The chatbot offers three distinct practice modes—Case Scenarios, Role-Play, and Improve a Provider’s Response (pictured)—allowing providers to strengthen different aspects of their counseling skills in a safe learning environment.
- Dynamic Client Personas: In Case Scenarios and Role-Play exercises, the chatbot takes on realistic client personas, responding naturally to the provider’s counseling while introducing key themes such as eligibility, self-injection fears, myths about contraception, and side effect discussions.
- Feedback Grounded in Training Materials: At the end of each exercise, the chatbot provides 2-3 concise feedback points drawn directly from evidence-based DMPA-SC training materials, reinforcing best practices in counseling.
- Multilingual Support: The chatbot operates in both English and French, ensuring accessibility for diverse users across francophone contexts.
- Structured Learning Pathways: Providers can choose their own learning journey, selecting exercises based on their specific needs and repeating exercises with new scenarios each time to build competency through varied practice.
Testing with Providers in Cotonou
In early 2025, Jhpiego facilitated a workshop with 14 healthcare providers in Cotonou, Benin, who shared their experiences and concerns raised by clients initiating (or wanting to initiate) DMPA-SC as a self-injected product.
These insights informed contextually relevant scenarios for the chatbot, and content was fed into OCS to refine the DMPA-SC coach further. Shortly thereafter, the healthcare providers were reconvened to test the chatbot and share their perspectives through individual surveys and group discussion, including role plays.
The chatbot was well received, with providers appreciating its practical application and immediate relevance to their work.
Key findings from the proof-of-concept:
- High satisfaction and perceived utility: 90% of participants were “very” or “completely” satisfied with the chatbot, and all participants felt the chatbot was useful to them and would be for others as well.
- Strong applicability: 90% of participants felt they could apply lessons from the chatbot immediately, and all agreed that it would help them manage stress of providing patient care on this topic.
- Accessibility for new users: Half of participants had never heard of LLMs like ChatGPT before the test, yet they were able to engage effectively with the tool, demonstrating its accessibility for providers regardless of prior AI experience.
- Language quality: The chatbot’s ability to communicate in French was very good, making it suitable for francophone contexts.
Positive Feedback:
“I am very pleased with this chatbot application and would like to see it made available online so that clients and their families can stay informed about everything related to the use of DMPA.”
“The chatbot must be active at all times. Expand the chatbot to include long-acting methods and respond to our concerns.”
”I would like to make the chatbot available to all service providers to enable them to acquire new knowledge and manage stress”.
Areas for Improvement:
- Maintaining scope and accuracy: While the chatbot mostly remained accurate and on-scope, there were some instances where it deviated from the intended counseling themes or training content.
What’s Up Next
The test proved the initial usability and perceived utility of the DMPA-SC Coach chatbot as a tool for strengthening provider counseling skills. Moving forward, there are promising opportunities to apply this and to explore how AI-powered solutions can further support contraceptive decision making and use.
At the same time we must ensure that AI-related applications in family planning are never coercive or reduce autonomy over decision making, are always accurate and reliable, that they protect patient data and confidentiality, and are equitably accessible.
The success of this proof-of-concept demonstrates that AI-powered coaching has the potential to address multiple touchpoints in the DMPA-SC adoption process, from strengthening provider skills to supporting client confidence and continuation, and suggests expansion to additional services that have a self-care or self-administered component is worth exploring.
By Megan Christofield, Jhpiego and Gayatri Jayal, Dimagi
